Analysis of green word-of-mouth advertising behavior of organic food consumers

The word-of-mouth (WOM) marketing process is one of the main means by which consumers obtain information. As a communication channel between consumers in economically developing countries, WOM may contribute to the development of the organic food market. The primary objective of this study is to segment organic saffron consumers in Mashhad, Iran, and determine how they engage in WOM marketing. Data were collected through questionnaires from 13 districts of Mashhad using a stratified sampling method. In this study, 400 organic saffron consumers were grouped using a self-organizing map (SOM) neural network based on consumer neobehavioristic theory, and then, using decision trees, consumer behavior rules were extracted for participating in the WOM for each group. According to the results, less than fifty percent of consumers in each of the four market segments are willing to participate in WOM advertising for organic saffron. A lack of awareness of the characteristics of organic saffron is also found to be the main reason for consumers ’ reluctance to recommend organic saffron to others. Mass-media advertising is an effective way to raise consumer awareness and influence opinion leaders, ultimately resulting in WOM recommendations.


Introduction
As a result of population growth and rapid economic development, the environmental condition has deteriorated, making this a global concern and a top priority (Wu & Chiang, 2023).The production of organic products and food is environmentally friendly and in line with sustainable development goals (Cachero-Martínez, 2020).Organic agriculture is an effective way to protect natural resources and the environment by using fewer chemicals, reducing soil pollution, avoiding harmful machinery for soil health, producing soil and plants properly, and reducing damage caused by pests and biological diseases (Esmaeilian et al., 2022).Past studies have found conflicting results regarding organic and conventional foods' nutritional properties; however, organic foods are believed to contain a wide variety of beneficial antioxidants (for more details, see Suciu et al., 2019).Organic products are healthier than conventional products due to their higher quality and the absence of chemicals and synthetic additives.Hence, organic products and foods are beneficial to both the consumer and the environment (Cachero-Martínez, 2020).Increasing life expectancy, demographic changes, and improving quality of life have made consumers more interested in eating organic food than ever before.Over the last two decades, demand for organic food has steadily and gradually grown to a global market size of USD 90 billion (Ali et al., 2021).From 2015 to 2020, the global organic food market experienced an annual growth rate of 16%.Although European and North American countries are the largest markets for organic foods, this market is expected to experience the highest growth rate in Asia over the next five years (Nafees et al., 2022).The growth rate of the organic food market, however, differs by country (Chakrabarti, 2010).Most consumers in economically developing countries have little or no experience with organic foods as the market is still small and underdeveloped (Chang & Chang, 2017).In the case of organic food and products, consumers are more at risk of making a wrong purchase decision, so they seek more information to reduce their perceived risk.For this reason, word-of-mouth (WOM) has become a valuable source of information and a crucial communication tool for green consumers (Li & Jaharuddin, 2021).
WOM has been one of the most influential channels of consumer information for the past half-century (Chang & Chang, 2017).Consumers frequently discuss the products, services, and marketing activities of companies (Martensen & Mouritsen, 2016).By enhancing customer awareness and facilitating information flow, WOM helps businesses increase their market share and gain an edge over their competitors (Konuk, 2019).It also allows companies to extend the impact of their marketing activities beyond their expected timeline and therefore influence consumer social behavior (Martensen & Mouritsen, 2016).It is believed that WOM communication via social networks and reference groups has a considerable impact on the organic food market (Chang & Chang, 2017).
The Razavi Khorasan province has more than 2000 ha of organic land, making it one of Iran's leading producers of organic products (Tohidi et al., 2023).There are more than 85,000 farmers cultivating saffron in the Razavi Khorasan province, the largest producer of saffron in Iran and the whole world.The plant plays an important role in the agriculture of the Razavi Khorasan province as a result of its less water consumption, higher productivity, rural development, and increased income.As one of the world's most valuable agricultural and medicinal products, saffron is widely used in the food industry and traditional medicine (Esmaeilian et al., 2022).Saffron is considered to be the most economically feasible and environmentally friendly product due to the climatic conditions of this province.The growing interest in organic saffron products as well as food safety concerns have led to the expansion of organic saffron production (Ghorbani & Koocheki, 2017).
A lack of consumer awareness is a major obstacle to the growth of the organic saffron market in Iran, which can be addressed through WOM campaigns (Tohidi et al., 2023).Several factors influence the intention to participate in WOM campaigns, but few studies have been conducted on these factors in economically developing countries such as Iran.There is also the problem that organic food consumers vary in terms of demographics and psychological characteristics, so the level of participation in green WOM advertising varies among different consumer groups, which has not been addressed in previous research.To examine the participation of organic food consumers in WOM advertising, most previous studies (e.g., Allen & Spialek, 2018;Bae et al., 2023;Hameed et al., 2022;Mansoor & Noor, 2019;Román-Augusto et al., 2023;Sun & Ham, 2022) have employed structural equations, which has led to a variety of different and somewhat conflicting results, as parametric statistical methods are highly dependent on model specifications.In general, statistical models are based on a top-down approach.These models assume that the model and the relationships between the variables are known and that the unknown parameters of the model are estimated from the data (Kori et al., 2023).In statistical models, the main problem is that the researcher must be aware of the relationships between the variables, and if these relationships do not match reality, the model will be suboptimal with inaccurate outcomes.The application of machine learning models is a bottom-up process.In machine learning methods, the model is not specified by the researcher but is created based on data and algorithms.As the parameters are adjusted during the training process, the output is obtained with the least amount of error possible.Therefore, machine learning methods are data-driven as opposed to statistical methods, which are chosen by the user (Ley et al., 2022).
This research fulfills an identified need to study which factors influence consumer intention to engage in green WOM advertising by answering the following questions: Is there a difference in participation in green WOM advertising among different segments of Iranian organic saffron consumers?What is the most important psychological and demographic characteristic that affects consumers' participation in WOM marketing?
This study uses the consumer neobehavioristic theory as a framework for answering research questions because previous studies (e.g., Hameed et al., 2022;Sultan et al., 2021;Tohidi et al., 2023) have demonstrated that taking external stimuli, psychological characteristics, and demographic factors into account, this theory is well suited to explain consumer behavior.According to the neobehavioristic theory, external stimuli (such as the marketing mix) induce behavioral responses in consumers by influencing their internal organisms (such as consumer attitudes) (Tohidi et al., 2023).The self-organizing map (SOM) and decision tree are also used to group, extract consumer behavior rules, and identify key factors for consumer participation in green WOM.The SOM is a type of unsupervised neural network and a non-parametric approach that is capable of classifying consumers into homogeneous groups without prior knowledge (Panda et al., 2022).Using an if-then rule set, a decision tree breaks a decision problem into sub-decisions for easier understanding and interpretation (Li et al., 2022).These two machine learning methods, which have rarely been used to study consumer behavior, are capable of identifying nonlinear and complex relationships between variables with minimal human intervention and without using simple and unrealistic assumptions.In fact, a novelty of this study is the use of computational intelligence in a framework of consumer neobehavioristic theory to gain a deeper understanding of Iranian consumers' participation in WOM advertising for organic saffron.

Literature review and theoretical background
According to Chang and Chang (2017), the term "green consumption" refers to consumers who tend to buy environmentally friendly products that reduce or minimize the environmental impact of their actions.Organic food consumption, green advertising and WOM, recycling, and buying environmentally friendly products are some of the topics related to green consumption in the literature.In comparison to conventional foods, organic foods are healthier and safer (Mesnage et al., 2020), as well as environmentally friendly (Chang & Chang, 2017).Organic agriculture and green consumption have received a lot of attention in recent decades (Carrión Bósquez et al., 2023), and numerous studies have been conducted on the subject (e.g., Issock et al., 2023;Liu et al., 2021;Nguyen & Dang, 2022;Richetin et al., 2022;Tandon et al., 2020;Yuan et al., 2024).The majority of organic food consumption studies have been conducted in economically developed and industrialized countries like Germany and the United States (for more details, see Kushwah et al., 2019).Due to the prevalence of WOM in today's society, it can have a significant effect on the development of the organic market in economically developing countries (Chang & Chang, 2017).
WOM communications, in marketing, refer to oral and person-toperson communications between senders and receivers that are considered non-commercial while the message's subject is usually about a product or brand (Chang & Chang, 2017).WOM is defined in another way as non-commercial, informal, and person-to-person communication between a sender and receiver about a product, service, brand, or organization (Konuk, 2019;Li & Jaharuddin, 2021).WOM communication may take the form of face-to-face conversations or online reviews (for more details, see Berger, 2014).Consumers find that WOM is the most effective way to share their opinions about products and services, and they consider that this information is more reliable and trustworthy than that obtained from commercial advertising (Li & Jaharuddin, 2021).By exchanging information, experiences, and opinions about a product or service, consumers can replace WOM with direct experiences and search for information effectively (Martensen & Mouritsen, 2016).However, WOM and its role in the development of the organic food market have often been studied in developed countries, but in economically developing countries, the issue has gotten less attention (Ali et al., 2021).The research literature has focused on factors that influence the consumption of organic foods (for more details, see Sharma et al., 2023), yet there is a lack of knowledge about what drives WOM recommendations for organic food consumption (Simanjuntak et al., 2023).Thus, the primary contribution of this study is to investigate factors affecting WOM intention among Iranian organic saffron consumers.
Methodologically, most studies have used structural equations, which is a parametric approach to explain consumer behavior related to WOM marketing.Several different and somewhat conflicting findings have been found regarding what affects WOM advertising, which may result from different model estimation specifications.The problem of F. Boccia and A. Tohidi misspecification, which is common in parametric models, results in systematic bias in their estimates and subsequently misleading conclusions (Yuan et al., 2003).In the real world, consumer behavior is complex and non-linear, and it is impossible to specify it using conventional and parametric models without making unrealistic assumptions (Tohidi et al., 2023).Artificial intelligence can make accurate predictions using machine learning algorithms to process large-scale and unstructured data.Machine learning methods have a flexible structure and strong performance in modeling consumer behavior by connecting computing power to human insight and marketing theories (Ma & Sun, 2020).The development of machine learning methods is based on this view, which enables the identification of hidden patterns in raw data with a minimum of human intervention.The use of machine learning methods provides accurate results for marketing studies; however, they have remained largely untapped for a wide range of consumer behavior studies (Tohidi et al., 2023).According to Vlačić et al. (2021), the field of marketing and artificial intelligence is new and nascent, and many applications of artificial intelligence in marketing studies remain unknown.Modeling and predicting consumer behavior is one of the main objectives of marketing science.By recognizing complex relationships among data and by recognizing consumer behaviors and attitudes, artificial intelligence can support marketers in designing marketing plans and making strategic decisions (Vlačić et al., 2021).Thus, the second contribution of this study is the use of machine learning methods to extract consumer behavioral rules regarding organic saffron WOM marketing.
According to the research literature, WOM communication studies can be divided into two general groups.The first group of studies examines WOM communication as a mediator and its impact on purchasing intentions (e.g., Kursan Milaković et al., 2020;Mansoor & Noor, 2019;Martensen & Mouritsen, 2016).WOM communication moderates the relationship between psychological variables and purchase intentions in this group of studies.The second group of studies investigates the factors affecting WOM as a dependent variable (e.g., Hwang & Kim, 2019;Issock Issock et al., 2020).Behavioral intention can be regarded as an important indicator of the consumer's intention to perform a specific behavior, which is influenced by attitudes and subjective norms (Chuang et al., 2017;Wu et al., 2015).WOM is considered to be one of the most reliable indicators of customer behavior (Kang et al., 2020), so it is used as a dependent variable in this study.Additionally, in terms of the neobehavioristic theory of consumer behavior, WOM is a relevant dependent variable.
To describe consumer behavior, there are two approaches: behavioristic and neobehavioristic.The first approach is based on the stimulus-response (S-R) framework, according to which external stimuli (such as the marketing mix) lead to a response from consumers (such as purchase behavior).According to this approach, the consumer is viewed as a black box, in which decisions result in a behavioral response based on the consumer's characteristics.The process inside the black box is invisible and is not addressed in research.The neobehavioristic approach, based on the stimulus-organism-response (S-O-R) framework, replaces the black box with the organism to explore the internal processes involved in decision-making.This approach is based on the assumption that internal organismic factors (such as consumer attitudes) can be measured through indicators (Rödiger & Hamm, 2015).According to Tohidi et al. (2023), the S-O-R framework by considering organismic factors or attitude variables can adequately explain consumer behavior for organic saffron.The S-O-R framework, according to Hameed et al. (2022), can identify the reasons behind consumer WOM behavior.In the S-O-R framework, it is assumed that stimuli (e.g., marketing mix elements) impact consumers' emotional conditions, which then lead to the formation of their behavior.The framework provides insights into the behavior of organic food consumers.According to this model, the formation of behavior (response) is influenced by both external environmental factors (stimuli) and the internal attitudes of the consumer (organism).In other words, behaviors are not only determined by external stimuli, but also by psychological characteristics.Based on the S-O-R framework, the behavioral response consists of three stages: (S) When the consumer is exposed to external stimuli, (O) a consumer's internal attitude is formed, and (R) which ultimately leads to a response or behavior (Hameed et al., 2022).
External stimuli are often able to influence consumer behavior through attitudes and psychological characteristics.As a post-purchase behavior, WOM is influenced by the psychological characteristics of the consumer (Boccia et al., 2013;Hameed et al., 2022).Recent studies (e.g., Schipmann-Schwarze & Hamm, 2020;Tohidi et al., 2023) have shown that incorporating demographic characteristics into the S-O-R framework improves its ability to describe green consumer behavior because these characteristics influence consumer behavior and preferences.
According to neobehavioristic theory and S-O-R structure, Fig. 1 illustrates the study's conceptual model.This figure shows the marketing mix as an external stimulus since it affects customer attitudes.Consumers' affective and cognitive conditions are formed by the organism, which includes all the processes between stimulus and response (Lavuri et al., 2023).Organism includes psychological factors such as knowledge, belief, adoption, and attitude toward organic food (Lavuri et al., 2023;Wang et al., 2022).Intention to participate in WOM advertising is influenced by acceptability (Martensen & Mouritsen, 2016), accessibility (Berger, 2014), affordability (Martensen & Mouritsen, 2016), awareness (Konuk, 2019), trust (Hameed et al., 2017), health consciousness (Chauke & Duh, 2019), and environmental concern (Issock Issock et al., 2020).Since these variables are the attitude and internal factors of consumers (Tohidi et al., 2023), they are considered organismic factors in this study.As a reactive behavior, WOM intention is influenced by both consumer demographic characteristics and organismic factors.Considering Fig. 1, in the following sections, the relationship between organismic factors and demographic characteristics with WOM intention will be discussed.

Marketing mix and word-of-mouth
As one of the most significant concepts in modern marketing, the marketing mix includes a set of practical marketing tools called 4Ps (i.e., product, place, price, and promotion).To achieve organizational goals and satisfy consumers, companies use marketing mix elements (Tohidi et al., 2023).A product is defined in terms of both its physical characteristics (quality, brand, features) and its service operations (warranty, after-sales service, complaint management, etc.).Place refers to the location and quality of the company's sales centers and website.Price is a key element, and it includes competitive prices, discounts, price lists, and value for money.Promotion refers to advertisements and general communication with consumers about the company's products (Martensen & Mouritsen, 2016).
The 4P marketing mix model has been criticized for focusing only on producer activities and ignoring the consumer perspective (Tohidi et al., 2023).Research on the relationship between marketing mix and WOM communication has been limited, and it has examined the relationship from the producer's (sender's) perspective rather than from the consumer's perspective (Martensen & Mouritsen, 2016).The 4P model can be described as the 4A framework (i.e., acceptability, accessibility, affordability, and awareness), where each P is evaluated from the perspective of the consumer (Tohidi et al., 2023).
Acceptability indicates the extent to which the product or service can meet consumer expectations (Tohidi et al., 2023).Consumer behavior research has investigated the impact of perceived product quality and service quality on customer satisfaction.As a result of the acceptability of a product, the consumer shares his/her experience with others and encourages them to buy it.It is most common for consumers to share their opinions about a product or service with others when the product is worth talking about and can meet their needs (Martensen & Mouritsen, 2016).It is believed that if a product is more acceptable to consumers, F. Boccia and A. Tohidi they are more likely to share their experience or discuss the product (Berger, 2014;Chauke & Duh, 2019).
A product's accessibility refers to how easy it is for consumers to purchase and use it (Tohidi et al., 2023).When consumers do not feel that a company's product or service is readily accessible or is not convenient, they are reluctant to recommend it (Martensen & Mouritsen, 2016).A high degree of visibility of a product makes it more accessible to consumers and is more likely to be discussed by them.In other words, accessibility acts as a key driver in the spread of WOM communication (Berger, 2014).
An external stimulus such as price can serve as an indicator of product quality in the consumer's mind (Boccia et al., 2023).The affordability of a product is determined by the consumer's ability and willingness to pay for it (Tohidi et al., 2023).In the literature (e.g., Issock Issock et al., 2020;Martensen & Mouritsen, 2016;Mladenović et al., 2021;Tohidi et al., 2023), it has been extensively discussed that price has a significant impact on consumer behavior when it comes to WOM advertising.The relationship between affordability and WOM, however, is not well understood from the consumer perspective.According to Martensen and Mouritsen (2016), when a product is able (unable) to meet consumer expectations and needs, WOM communication increases (decreases).
Consumer awareness refers to the extent to which consumers know about the benefits and features of a product (Tohidi et al., 2023).By increasing awareness and facilitating the flow of information among consumers, WOM helps companies increase their market share and gain a competitive advantage (Konuk, 2019).WOM is more likely to spread when people know more about a certain product because it allows them to demonstrate their knowledge and expertise (Berger, 2014).Although research results (e.g., Konuk, 2019;Martensen & Mouritsen, 2016) suggest that awareness can influence WOM communication, there is little evidence to support this claim.

Trust and word-of-mouth
Trust can be defined as the anticipation created by consumers in response to the level of fulfillment of promises made by product suppliers.Consumers have little capacity to evaluate the quality of organic products, so trust is a key advantage in the development of the organic products market.It is believed that trust has a significant impact on consumers' behavioral intentions.Accordingly, consumers are more likely to act on WOM recommendations when they are more confident about organic product labels and certifications (Cachero-Martínez, 2020; Issock Issock et al., 2020).Hameed et al. (2017) found that customer WOM intentions are often influenced by trust.

Health consciousness and word-of-mouth
Health consciousness is one of the factors influencing organic food consumers' behavior, as it relates to their attitude toward health-related decisions and activities (Ali et al., 2021).Therefore, an important aspect of health consciousness is the readiness to engage in health-promoting activities (Li & Jaharuddin, 2021;Covino and Boccia, 2013).According to Jin et al. (2017), health consciousness plays an important role in people's behavior toward eating healthy food and spreading WOM recommendations.A study conducted by Chauke and Duh (2019) found that people are more likely to talk about their food choices when they are concerned about their family's health.

Environmental concern and word-of-mouth
The production of organic food is environmentally friendly, and it contributes to sustainable development.Consumers' environmental concern is defined as their awareness of environmental issues and willingness to solve them.Increased environmental concerns lead to an increase in consumer attitudes and willingness to spread WOM (Cachero-Martínez, 2020).According to Issock Issock et al. (2020), consumers will participate in WOM advertising if they are confident that the organic product will comply with environmental standards.

Demographic characteristics and word-of-mouth
Demographic characteristics (gender, age, education, and income) are believed to play a significant role in shaping consumer behavior toward organic food in economically developing countries.A study by Tohidi et al. (2023) found that demographic factors are closely related to organic food consumer behaviors.Previous studies indicate that demographic factors influence WOM intention.Evidence suggests that F. Boccia and A. Tohidi gender is a significant factor in WOM engagement.Women and men participate in WOM communication for different reasons.The reason men use WOM communication is to maintain their social status, whereas women use it for cooperation and emotional reasons (Kursan Milaković et al., 2020;Mladenović et al., 2021).Hwang and Kim (2019) also found that gender influences eco-friendly behaviors and WOM intentions.Additionally, age and education have been shown to have a significant influence on WOM intentions (Kursan Milaković et al., 2020;Mladenović et al., 2021).Education is an important factor in choosing healthy food as it improves the knowledge and attitude of the consumer (Boccia et al., 2023).People with high levels of education have more intention to spread WOM because they have more information to share (Mladenović et al., 2021).Hwang and Kim (2019) concluded that older people are more likely to participate in WOM because of concerns related to green issues.Tohidi et al. (2023) argued, however, that young people are the target group for advertising campaigns in economically developing countries.Although it is believed that income does not have a significant effect on WOM communication, the results of studies show that people with higher incomes are more likely to participate in WOM (Kursan Milaković et al., 2020).
In general, few empirical studies have examined the effect of demographic variables on WOM intention, and the analyses have relied on generalizations, which is a major research gap (Mladenović et al., 2021).F. Boccia and A. Tohidi

Study area
Mashhad is among the largest cities in Iran and has a history dating back 1200 years.It occupies an area of approximately 290 km 2 and is the capital of Razavi Khorasan province (see Fig. 2).There are 13 districts in Mashhad, which is the second most populous city in Iran after Tehran, with 930 thousand households and three million residents (Tohidi et al., 2023).Historically, Mashhad has been one of the largest producers and suppliers of saffron, a product that plays a significant role in small farmers' economies (Ghorbani & Koocheki, 2017).
Due to its medicinal and nutritional properties, saffron is a staple spice in the food baskets of residents of Mashhad (Tohidi et al., 2023).Considering Mashhad is the main saffron market in Iran, it was selected as the study area.

Measures and data collection
This study utilizes the data collected from Tohidi et al. (2023), with the difference being that the behavioral variable is WOM intention.Based on a stratified sampling (with proportionate allocation) from 13 districts of Mashhad, the sample data includes the opinions of 400 respondents, who were interviewed in January and February 2020 using face-to-face questionnaires (for more details, see Tohidi et al., 2023).This sampling approach has the advantage of including a population representative of each of the Mashhad districts in the statistical sample.According to the conceptual model (Fig. 1), WOM intention, as a dependent variable, is influenced by both organismic variables and demographic characteristics.The organic saffron WOM intention is a binary variable with a value of 1 indicating that the respondent is willing to participate in WOM, and a value of 0 otherwise.According to the S-O-R framework, the elements of the marketing mix, trust, environmental concern, and health consciousness are the organismic variables that influence WOM intention (see Table 1).The organismic factors in Table 1 are scored using a scale from 1 to 10 in accordance with previous studies (Bernabéu et al., 2018;Hempel & Hamm, 2016), where a score of 1 indicates no extent and a score of 10 indicates a large extent.The effect of demographic characteristics, i.e., age, gender, income, and education, on respondents' intentions to participate in WOM for organic saffron has also been explored in this study.

Self-organizing map
Due to the diversity of organic food consumers' needs, desires, demographic characteristics, motivations, and behaviors, a single model cannot capture behavioral patterns for all consumer groups.Accordingly, in this study, consumers are divided into homogeneous groups based on the variables in Table 1, and then their WOM participation factors are extracted for each group.A type of artificial neural network (ANN) is therefore used in the present research to group organic saffron consumers.
In recent years, ANNs have gained considerable attention due to their capability to solve a wide range of business problems (Panda et al., 2022).The purpose of ANNs is to provide solutions to problems with a function similar to that of the human brain.Neurons or nodes are one of the main components of an ANN whose functions are derived from biological neurons.ANNs perform better when trained properly as opposed to econometric methods that rely on the validity of statistical assumptions.Consequently, learning or training plays a significant role in the implementation of an ANN.There are two types of ANNs, supervised learning, and unsupervised learning, depending on the training method.Supervised learning is a method in which an ANN is trained on two sets of inputs and targets.During this process, the input weights are adjusted to minimize the error between the ANN's output and the real value.In unsupervised learning, the desired response is unknown, so the error cannot be calculated to improve the ANN's performance.All variables are considered input variables in an unsupervised method since there is no target variable.The output neurons compete with one another to match the multidimensional space of the training data (Benbrahim Ansari, 2021).Several applications of unsupervised ANNs are found in the fields of clustering, visualization, and exploratory research.Using SOM, a type of unsupervised ANN, high-volume data can be represented in a two-dimensional space.It can be used to specify topology or relationships between data (instead of actual distances) on a map.Since SOM is a non-parametric approach that requires minimal prior knowledge and preprocessing of data, it is superior to statistical and optimization methods (Panda et al., 2022).In addition, by adding new data, it is possible to determine which node the data belongs to (Kim, 2022).In comparison to other regression and classification methods, SOM exhibits greater flexibility, which has led researchers to employ it in applied research.The use of this method has, however, been very rare in studies related to consumer behavior.Through the recognition of similarities among features used for clustering, the SOM algorithm can identify and display the topological structure hidden in the input data (Panda et al., 2022).
The data structure is preserved by assigning similar input vectors to adjacent output neurons in low-dimensional output space.A competitive and cooperative learning approach is an essential feature of the SOM method.In competitive learning, a sample vector is randomly selected from the input data set, and then the Euclidean distance between the input vector and the output neurons is calculated.As a result, the best matching unit (BMU) is the neuron whose weight vector is closest to the sample vector.In cooperative learning, both the BMU and its neighbors' weight vectors are updated to better match input vectors (Alkahtani et al., 2019).
The following is a description of the SOM's algorithm (Kim, 2022).F. Boccia and A. Tohidi 1.To initialize the neuron value, data covariance Ω is disassembled into its main components, from which the two principal components with the highest descriptive power are extracted.2. The data x is compared with all n i (i = 1, 2, …, N) neurons in the ith iteration to determine neuron n b (i.e., BMU): 3. A recalculation of the attribute neuron n i is performed using the mean of the aggregated data: where the index j corresponds to the vectors of input data for neuron b, and K represents the number of input data vectors.The neighborhood or Gaussian function h ic(j) is defined by equation ( 3): Where the numerator represents the square of the Euclidean distance between reference neurons n i and n j in a self-organizing map.The parameter σ represents the radius of the neighborhood at time t, which decreases as a function of time throughout the algorithm.
4. Steps 2 and 3 are repeated a number of times as specified.

Decision tree
Following the segmentation of the respondents, a decision tree is used to extract their behavioral rules related to the WOM of organic saffron in each of the segments.Human logic has been applied to develop the decision tree technique, one of the most powerful machine learning techniques (Chi-Hsien & Nagasawa, 2019).It should be noted that decision trees are not only used in machine learning and data mining, but they also hold a significant place in statistics and can be used in multiple fields of behavioral science as well.The decision tree has been developed independently in statistics and machine learning fields and is widely used and popular for learning discriminant models.This technique is commonly used for classification applications, and it is relatively easy to understand (Kliegr et al., 2020).
A new instance is classified at the root, which is the top node.In the following step, the classification process proceeds downward by moving from the branch with the determined value of an attribute to a new decision node containing the new attribute.As the process progresses, it reaches the final leaf or node of the tree.Top-down learning is used in decision trees: the best feature is selected for the tree's root, followed by the addition of its nodes and branches (Kliegr et al., 2020).It focuses primarily on those attributes that are most important in determining the target variable (Chi-Hsien & Nagasawa, 2019).
The interpretation of a decision tree is based on if-then rules.The logic rules used in a decision tree are similar to those used by humans when making decisions, which makes it easier to understand and interpret.Using a set of logical rules, the decision tree combines a series of simple tests to break down a problem and fit a model into each subproblem.A decision tree is a graphical representation of the results of the models.The nonparametric nature of this model allows it to be applied to a wide range of data with a variety of statistical probability distributions.Researchers can use this advantage to model many problems in which the relationships between variables are unknown.As well, the decision tree requires no data transformation, which makes it easier to use.Due to these advantages, decision trees have been applied widely for the analysis of regression, classification, and prediction in the data science field.Decision trees have been developed with numerous algorithms with varying features and capabilities; however, they are statistically and practically equivalent in their accuracy.Depending on whether the target variable has binary or continuous values, classification decision trees and regression decision trees are used, respectively (Li et al., 2022).As the intention to participate in WOM for organic saffron is a binary variable, classification trees are used in this study.

SOM-based segmentation
By implementing a network with 12 neurons in the input layer, which corresponds to the number of segmentation criteria, the SOM method is used to group organic saffron consumers.Thus, the input dataset consists of a matrix of 12 by 400, representing 12 segmentation criteria and 400 consumers (see Table 1).The BMU is determined by selecting a sample vector randomly from the input data set and following the minimum distance rule.SOM plot sample hits in Fig. 3 represent neuron locations in the topology as well as the number of organic saffron consumers within each cluster.The output layer is a twodimensional N × N space, where N represents the number of neurons.According to Fig. 3, consumers are grouped into four market segments using two neurons in the output layer.Table 2 summarizes the characteristics of the respondents in each of these groups.
Table 2 indicates that the first segment of the market is dominated by females, highly educated and higher-income consumers.In this segment, the average score for organismic factors is higher than in the others.Furthermore, consumers of organic saffron in the first segment give the highest score to health consciousness among organismic factors.As compared to the other three segments of the market, the first segment has the highest percentage of respondents intending to participate in WOM marketing.
In the fourth segment of the market, organic saffron consumers have the least intention of participating in WOM marketing.The lowest scores were given to organismic factors by consumers with relatively lower incomes and levels of education.
Among the segments of the organic saffron market, the second and third segments are the oldest and youngest, respectively.Over half of the sample population belongs to these two segments.These two segments of the market have lower scores for organismic factors than the first segment and higher scores than the fourth segment.However, the Fig. 3. SOM sample hits.
F. Boccia and A. Tohidi percentage of people intending to participate in WOM is higher in the third market segment than in the second market segment.

Decision tree-based extraction of decision rules
In this study, classification decision trees are used to determine behavioral rules that influence organic saffron consumers' WOM marketing participation.Fig. 4 illustrates what factors influence consumers' intentions to participate or not participate in WOM advertising for organic saffron in the first market segment.Furthermore, the decision tree shown in Fig. 4 represents the threshold values for each of the key variables where bifurcation has occurred.
Consumer intention to engage in WOM is most strongly influenced by environmental concern, age, income, accessibility, and health consciousness in the first segment of the market.There are three ways in which consumers intend to engage in WOM advertising in this market segment: (1) environmental concern greater than 4 and age greater than 38, (2) environmental concern greater than 4, age less than 38, and income greater than 3.25, (3) environmental concern greater than 4, age less than 38 and income less than 3.25, accessibility greater than 4 and health consciousness greater than 6.In contrast, consumers have less intention to participate in WOM advertising if the environmental concern is less than 4, accessibility is less than 4, and health consciousness is less than 6.
In the second segment of the market, WOM intention is strongly influenced by awareness and income, as shown in Fig. 5.In this segment of the market, WOM behavior is expected to occur when consumer awareness is greater than 4.5 and income is greater than 3.9.
According to Fig. 6, in the third segment of the market, awareness, education, affordability, acceptability, and health consciousness play a significant role in influencing WOM recommendations.Accordingly, the higher the value of these variables, the higher the likelihood that consumers will recommend organic saffron to others.Fig. 7 illustrates that, in the fourth segment of the organic saffron market, two key variables can explain consumer behavior, namely, affordability and awareness.

Discussion
It is believed that WOM communication is one of the most significant factors contributing to the development of the organic food market in economically developing countries.In all market segments, however, less than half of Iranian consumers are willing to participate in WOM advertising for organic saffron, according to the results of the SOM method.The fourth segment of the market only has about four percent of consumers engaging in WOM marketing, which is a low level of participation, despite saffron being one of the most popular spices in Iranian households.Several studies have shown that the extent of WOM communication varies depending on the product and subject matter.Allsop et al. (2007), for example, showed that people are more likely to share information about computers than about healthy eating.Opinion leadership, on the other hand, is believed to have a significant impact on WOM communication (Mladenović et al., 2021).There is a belief that opinion leaders and consumers engaged in WOM marketing have different personality traits compared to other consumers (Mothersbaugh et al., 2019).In the first segment of the market, the decision tree indicated that environmental concern has the greatest impact on the  F. Boccia and A. Tohidi willingness of organic saffron consumers to participate in WOM marketing.Consumers have grown more concerned about environmental issues over the past few decades due to media reports, industrial pollution, and the growth of environmental protection groups (Mansoor & Noor, 2019).This leads consumers to buy and recommend environmentally friendly products.It is therefore reasonably common to find a significant relationship between pro-environmental attitudes and green WOM advertising (Allen & Spialek, 2018).Studies by Cachero-Martínez (2020), Allen and Spialek (2018) and Covino et al. (2013) confirm the present study's findings, showing that consumers' intentions to engage in WOM increase with environmental concern.
The results indicate that awareness plays the greatest role in WOM advertising in the second and third segments of the market.Consequently, organic saffron consumers are more likely to participate in WOM advertising as their awareness increases.According to Mansoor and Noor (2019), consumers who are more aware of green products are more likely to participate in WOM advertising and persuade other consumers to purchase green products.Knowledge and awareness of products on the market generally lead to a greater willingness to share information among consumers (Berger, 2014).According to the Multi-step flow of communication theory, opinion leaders and consumers engaged in WOM marketing obtain information about the market directly from the mass media and then share it with other consumers (Schiffman et al., 2012).One of the reasons for low consumer awareness   F. Boccia and A. Tohidi of organic saffron in the Iranian market is the lack of advertising for organic saffron in the mass media.According to Tohidi et al. (2023) and Aghasafari et al. (2020), one of the main obstacles to the development of organic saffron in Iran is a lack of consumer awareness of this product's characteristics.Another consequence of the lack of awareness of organic saffron consumers is their unwillingness to engage in WOM advertising.
Results of the study indicate that affordability is more significant than awareness in influencing WOM marketing in the fourth market segment.Based on the results of the SOM method, the average income for this segment of the market is lower than for other segments, which explains why the price is the most critical factor in determining whether to participate in WOM marketing.Consumers are believed to be more willing to engage in WOM marketing if they perceive the price of the organic product to be reasonable and acceptable (Konuk, 2019).A study conducted by Martensen and Mouritsen (2016) found that affordable prices may have a positive impact on WOM marketing.
There is some evidence that gender plays a significant role in shaping WOM communication in some societies, and the type of communication is influenced by gender differences (Mladenović et al., 2021;Covino and Boccia, 2016).However, based on the findings of the present study, gender is the only variable that does not significantly impact WOM participation in any market segment.The results of a study by Sun and Ham (2022) are in line with the findings of this study, showing that gender influences psychological aspects such as personality traits and emotions, while it does not affect cognitive aspects such as WOM intention.

Conclusions
In economically developing countries such as Iran, a lack of awareness is one of the main barriers to organic food market development.WOM plays an important role in increasing consumer awareness of organic food as it is one of the primary channels by which consumers obtain information about organic food.However, there is little empirical evidence concerning factors influencing organic food consumers' willingness to engage in WOM advertising in economically developing countries.In this study, based on the S-O-R theory and using SOM and decision tree methods, the organic saffron market was segmented and consumers' WOM behavior was examined in each segment.SOM results showed that consumers can be grouped into four segments of the market so that they are homogeneous in terms of their demographic characteristics, organismic factors, and WOM behavior.It was found that less than fifty percent of consumers intend to participate in WOM advertising for organic saffron across all market segments.Two general conclusions can be drawn in response to the research questions.First, due to the different structure of decision trees within each of the market segments, it can be concluded that consumer participation in WOM advertising differs between the market segments owing to the heterogeneity of consumers' psychological and demographic characteristics.Second, psychological factors have a greater impact on consumers' participation in WOM advertising than demographic characteristics.WOM participation in organic saffron is strongly affected by environmental concern, awareness, and affordability, according to the decision trees.

Theoretical and practical implications
Results of the study have indicated that consumers' intention to participate in WOM advertising varies by market segment.From a theoretical perspective, to account for heterogeneity among consumers engaged in green WOM advertising, it is necessary to incorporate market segmentation into conceptual models.
As the identification of advertisers is the first and most critical step in creating a WOM campaign, the results of this study may be useful to marketers and policymakers when formulating WOM marketing strategies and developing advertising campaigns.More than half of the population of the organic saffron market is located in the second and third segments of the market.A characteristic of these segments is the low level of consumer awareness of the characteristics of organic saffron.Considering awareness plays a significant role in WOM advertising in the second and third segments of the market, it is plausible to conclude that consumers are not participating in WOM marketing for organic saffron due to a lack of awareness.Consequently, Iranian saffron manufacturers should increase consumer awareness of the characteristics of organic saffron in advertising campaigns targeting the second and third segments of the market.Accordingly, Iranian firms supplying saffron should focus on increasing consumer awareness of organic saffron's characteristics in advertising campaigns related to the second and third sectors.Furthermore, increasing advertising in the mass media is another effective strategy to enhance the flow of information between opinion leaders and organic saffron consumers.
Advertising programs should emphasize organic saffron's environmentally friendly production process in the first segment of the market, which is characterized by a relatively high level of environmental concern.For consumers in the fourth segment of the market, which is characterized by a low level of income and education, affordability plays the largest role in their intention to participate in WOM advertising.As a result, organic saffron promotional programs for consumers in this market segment should emphasize the distinctive features of organic saffron to encourage them to pay a premium price for it.
In this study, two machine learning approaches, which have been less widely used in previous studies related to consumer behavior, were applied practically.In today's world, agricultural companies can use artificial intelligence-based technologies to understand consumer behavior more accurately and effectively with the expansion of machine learning algorithms.

Limitations and further research
Data collection through a face-to-face questionnaire was challenged by time and cost constraints.Because of the limited sample size, organic saffron consumers were grouped into four distinct segments to provide enough samples in each segment.To have a more accurate understanding of consumer behavior, it is recommended that future studies implement market segmentation and the extraction of consumer behavioral rules for larger samples.Each element of the marketing mix can include several items.Due to time and cost constraints, these items are ignored in this manuscript.A more accurate marketing mix indicator can be created in future studies by taking into account a variety of items.
As consumers have varying interests and attitudes, market segmentation is crucial when studying green consumer behavior.Also, Decoding consumer perceptions with machine learning methods is one of the interesting fields in market research.There are several advantages to machine learning methods over statistical methods, including their flexibility, accuracy, and computational power.Machine learning methods used in this study provide a deeper understanding of green consumer behavior and can be applied to other areas of marketing research.
This study was conducted to investigate the behavior of organic saffron consumers since saffron is a highly consumed food item in Iranian households and plays an important role in the economy of Razavi Khorasan province.In future studies, it is suggested that consumers' behavior to participate in green WOM be investigated for other products and geographical areas.
A crucial aspect of WOM advertising is the credibility of the person promoting the product.It is suggested that future studies take into account this crucial aspect of WOM marketing.It is important to note that WOM communication is dynamic in nature.In fact, communication between consumers affects their attitudes and behavioral responses.It is therefore suggested that future studies investigate the dynamic effects of WOM communication on the behavior of organic food consumers.A more comprehensive understanding of WOM advertising behavior can be gained by adding lifestyle, social class, and personality traits to the F. Boccia and A. Tohidi consumer's neobehavioristic model.

Ethical statement
The study and the survey obtained ethics approval by the independent IRB of the authors' institution (Regional Research Centre, ethics/ approval number prot.n• 0421).Then, the participants gave informed consent before taking part.Children and youth were not allowed to participate in the survey.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Table 1
Study variables.

Table 2
Characteristics of market segments.