Comparative study of Jrip j48 and naive bayes algorithm in Flower specie prediction

Artificial intelligence and Deep learning techniques propose and provide effective mechanism for classification among several commodities like gender classification on basis of ridge count, spam mail classification and detection etc. Here we have proposed a hybrid module on basis of available AI and ML techniques by which we can achieve more than 90% accuracy for any provided test dataset. We have included over 50423 samples, out of which we use 66 percentage of data for training purpose of our model and 34% remaining sample we use as test dataset, with 8 fold we have achieve approximately 96% accuracy. For collection of dataset and processing of dataset we gone through several phases which include extraction of feature (EoF) using feature extraction technique, for cleaning of dataset we have use dimension reduction technique factor analysis(FA). In next phase for classification of flowers we have used classification techniques in which we found the accuracy score of j48 classifier is higher than Naïve Bayes and Jrip algorithm. Hence the choice of j48 classifier for classification is right.


Introduction:
Now a day's image classification is rapidly increasing its area of being in interest of data analyst and data scientists and the classification results and accuracy of results is giving effective mode of When we talk about Google photos, it can be considered as best example of image processing and classification as it correctly classify and group photos of person with higher accuracy along with that it take suggestion from user about similar images. With Google photos we can easily check our images with additional search (Triguero & Vens, 2016, p. 171) (Mukane, 2013, p. 83).
In this world over millions of species of flower exist if we wanted to make a model by which we can classify the species of flower with accuracy then AI and ML algorithms comes into picture, as the algorithm with its ability to classify a class or specie of flower with provided details about that flower leads to produce results of higher accuracy. Bloom order frameworks(Abdul Aziz et al., 2019a, p. 25) (Liu, 2014, p. 285)have a place with a shape coordinating issue, which is a major issue in PC vision and example acknowledgment. It characterized as the foundation of a similitude measure among shapes and its utilization for shape examination (Mukane, 2013, p. 83) (Abdul Aziz et al., 2019b, p. 30). The significant hypothetical fascinating inspiration for picture grouping comes from that shape coordinating is instinctively precise for people and need an associate which isn't fathomed at this point in its full consensus. Shape coordinating framework incorporates object acknowledgment and location, picture attachment, and substance based recovery of pictures (Mukane, 2013, p. 83).

Algorithm Used:
2.1 J48: J48 algorithm comes under Weka platform in a inbuilt manner which is basically implementation of C4.5 algorithm which generally used for making decision in a hierarchical manner, are known as decision trees, and when we combine two or more algorithm to make decisions then that combination is known as random forest(Koul & Singhania, 2020, p. 1726). C4.5 algorithm was a advancement of original ID.3 algorithm whose limitations later on overcome in C4.5 algorithm. Whatever decisions made by these algorithms (in a hierarchy) (   . Blossom locale can be divided dependent on utilizing shading highlights, since its pictures comprise of a huge green territory and bloom zone. The green region speaks to the leaves encompassing the bloom and blossom territory speaks to the bloom area which is portrayed by its tone ((Jain, 2017, p. 89). Shading fragmented can be accomplished utilizing the distinction/distance between two tones. In this work, the pictures are changed to Lab shading space. At that point, a casing which incorporates the bloom is chosen by client. The Delta E is utilized to compute the distance between each pixel in the chose outline and the normal LAB tone (Jain, 2017, p. 49 ). From that point forward, OTSU limit is applied on every Lab tone to fragmented bloom. OTSU calculation attempts to locate an ideal division between classes by registering a worldwide edge for a picture (Sharma, 2017, p. 83)

Experimental Results
Our experimental approach has been carried out to evaluate the proposed Hybrid module of 2 types of classifiers to have comparative analysis of two kinds. The proposed system was implemented using WEKA 3.8.3 on Windows 8.1 operation system. The hybrid combination is evaluated using around 50423 samples which represented 3 species. The used dataset in this experiment has been taken from Git hub in which we have add additional entries via extraction of feature f each and every specie and created around 50423 sample after that we have use 66 present of data as training data for our proposed model and remaining data which was chosen at complete random as test data. The features which we have considered in this dataset are: Saple-width: We have taken basic measurement in which flower of each species may have Saple-Height: We have taken basic measurement in which flower of each species may have Petal-width: We have taken basic measurement in which flower of each species may have Petal Height: We have taken basic measurement in which flower of each specie may have Specie: Specie is our class variable or we and say target variable. Target variable is always a dependant variable which depends on independent variable which are our other features which we consider as variable which play significant role in prediction of target variable      Figure 11: Naïve Bayes Classification output-III

Conclusions And Future Work
A hybrid approach is proposed which is capable to identify the species of flower which is provided as data to be tested with approx. 96 present accuracy. The proposed approach incorporates three stages: division, highlight extraction, and arrangement stages. Division stage plans to improve the exactness by isolating the blossom shape from the picture normal foundation. After that with chosen classification algorithm we have applied each algorithm on prepared dataset and with the comparative analysis we have found that The tree classifier and rule based classification algorithm shows best accuracy measure in prediction of correct specie i.e. Setosa, Veriginica, vericolor with the accuracy of 100%, whereas selected Naïve Bayes algorithm shows good results but in comparison of rule based and tree classifier it shows 96% accuracy. This hybrid model of combination of three algorithms will be suitable of any kind of flower specie and with some modification it can exceed its limitation to category other than flower also. This hybrid model shows nearly maximum possible accuracy on such a large dataset and with number of folds it won't take much time in calculation of correct specie, with these characteristics of hybrid module we are proposing this Hybrid combination of algorithm for data analysis and prediction purpose for future also. .