SYSTEMATIC ANALYSIS OF BLUE-CHIP COMPANIES OF NIFTY 50 INDEX FOR PREDICTING THE STOCK MARKET MOVEMENTS USING ANFIS MACHINE LEARNING APPROACH

Stock market plays an important role in the capital formation of a country. The stock market is often considered as the primary indicator of a country’s economic condition, its strength, and development. Companies with a good performance supposedly will have a good demand on its stock, hence boost the price and vice versa. However, there’s manipulation game in the market. Rumors, speculation and short-selling are among the manipulation activities that affect the fluctuation of stock price. The present study is intended to analyse the risk and return of select blue chip companies listed in NIFTY 50 Index using Adaptive neuro-fuzzy inference system, which may prove to be beneficial to the investors who makes investment.


INTRODUCTION AND PROBLEM DISCUSSION
With the increasing global competition, companies are focusing their efforts on creating shareholder value in order to survive the intense competition. In view of this, it is becoming important for companies to measure the value they create for their shareholders. Keeping track of the value created year-on-year enables companies to evaluate past decisions and make decisions that will improve shareholder value. Investors and market analysts resort to financial statement analysis when it comes to share investing. The information on Earnings per Share (EPS) is presented on the Income Statement while Return on Assets (ROA), which is one of the profitability ratios, is computed using relevant numbers from the Income Statement and Balance Sheet. The broad area of financial accounting and reporting offers a number of fundamental measures of a firm's performance for a particular accounting period.
Financial performance announcement is the process of producing statements that disclose an organization's financial status to the internal and external users of the information. The four basic reports are balance sheets, income statements, cash flow statements and statements of shareholders' equity. Financial reports are often prepared to cover a specific financial period.
While some cover an entire year, other reports only cover a specific period, such as a quarterly report. Financial statements act as tools of communication on a firms' performance especially during the period under review, performance announcement provide a modest but not overwhelming amount of information to the market. Fama (1970) states that in an efficient market, the security prices are presumed to reflect the effects of information based on past, current and future events. In an efficient market, any earnings or dividend announcement contains information which influences the stock prices positively or negatively. Investors use that information to make investment decision accordingly. However, the validity of the efficient markets hypothesis has been questioned through the emergence of behavioural finance by postulating that financial markets might fail to reflect economic fundamentals under a number of conditions which can result in significant and persistent biases. Hirshleifer (2001) explained that investors are not always rational and may not correctly process all available information while 267 SYSTEMATIC ANALYSIS OF BLUE-CHIP COMPANIES forming their expectations of an asset's future performance and, as such, trades could occur as a result of such irrationality.
On this, the researcher has structured the following research objective: • To analyse the risk and return of select blue chip companies listed in NIFTY 50 Index.

Qualitative Analysis
News feeds regarding stock market highly affect the market trend and thus form a downhill movement in case of negative news. Thus, the media/social network and stock market data are highly coupled and make the system more unpredictable. Existing research points out that in case of crisis, stocks mimic each other and lead to market crashes (Hellstrom 1998). Nowadays, Twitter has come forth as the most reliable and fastest way of consuming media. With combined resources of news feed and Twitter feed, general population sentiment about a company can be highlighted. Text mining and sentiment analysis are useful tools for such a high-scale analysis.

Quantitative Analysis
Historical data is now readily available for most markets. Using this dataset, we can apply multiple machine learning models to give accurate results for future investments. These models can be trained for individual stocks with adjusted bias for most reflective features. These models can also be trained to work in different scenarios and overall market movement Traditional approach focuses on fundamental analysis and technical analysis to predict the market at a large scale which rarely translates to low-level individual Stock Prediction, but it can be clearly observed that individual stocks contribute to whole market movement rather than the other way around. Thus, focusing on individual stocks to predict market movement is a much more logical approach. With technology advancing at such a rapid pace and abundance of computing power, we can now easily strive towards a comprehensive system to accurately predict the market trend and reap beneficial financial returns. Existing research proves that modern approach outperforms traditional approach and can output the most accurate results (Hellstrom 1998).

METHODOLOGY OF THE STUDY
Research has been conducted using data that are mostly secondary in nature and they are

Performance Criteria
To compare the classification models and evaluate their performance, three different performance criteria are used: accuracy, sensitivity, and specificity. Computation of these performance metrics are adopted as follows:  On the other hand, if it is predicted to fall, this signals a "sell". On the first day that a buy signal is received from the model, an investment is made in the BIST 100 fund using the full capital amount. Buy signals on subsequent days are treated as hold instructions until a sell signal is received. All investment units currently held are sold at the current price after a sell signal, and the capital is converted back to cash. Subsequent days' sell signals are again treated as non-trade instructions, until another buy signal is received, and the process is repeated.

ADAPTIVE NEURO-FUZZY INFERENCE SYSTEMS FOR FORECASTING
Adaptive neuro-fuzzy inference system is systematic estimation model not only among neuro-fuzzy systems but also various other machine learning techniques. The Adaptive Neuro-Fuzzy Inference System technique was originally presented by Jang (1993). ANFIS is a simple data learning technique that uses Fuzzy Logic to transform given inputs into a desired output through highly interconnected Neural Network processing elements and information connections, which are weighted to map the numerical inputs into an output.

ANFIS combines the benefits of the two machine learning techniques (Fuzzy Logic and Neural
Network) into a single technique, Jang (1993). An ANFIS works by applying Neural Network 271 SYSTEMATIC ANALYSIS OF BLUE-CHIP COMPANIES learning methods to tune the parameters of a Fuzzy Inference System (FIS). There are several features that enable ANFIS to achieve great success Jang (1995Jang ( , 1997): • It refines fuzzy IF-THEN rules to describe the behavior of a complex system; • It does not require prior human expertise; • It is easy to implement; • It enables fast and accurate learning; where: • $x$ and $y$ are the inputs, • $A_i$ and $B_i$ are the fuzzy sets, • $f_i$ are the outputs within the fuzzy region specified by the fuzzy rule, and • $p_i$ , $q_i$ , and $r_i$ are the design parameters that are determined during the training process.

Fig. 2. ANFIS architecture
The ANFIS architecture used to implement these two rules is shown in Fig. 4. In this figure, a circle indicates a fixed node, whereas a square indicates an adaptive node. ANFIS has a five-layer architecture. Each layer is explained in detail below.
where $x$ and $y$ are the inputs to node $i$ , and $A_i$ and $B_i$ are the linguistic labels is employed, ( ) is given byz or the Gaussian membership function by where $a_i$ , $b_i$ , and $c_i$ are the parameters of the membership function.
In ${\rm Layer}_{(2)}$ , the nodes are fixed nodes. This layer involves fuzzy operators; it uses the AND operator to fuzzify the inputs. They are labeled with $\pi$ , indicating that they perform as a simple multiplier. The output of this layer can be represented as where $\bar{w}$ is the output of ${\rm Layer}_{ (3)  have Beta more than 1. Hindustan Unilever Limited has the highest beta of 2.11. Though these stocks with beta greater than 1 are aggressive securities but stock of Hindustan Unilever Limited is most volatile of all the stocks. The returns of these companies are more volatile to changes in the market. Some securities like Bharti Airtel (0.68), HDFC Bank (0.79), HDFC Ltd (0.88) and ITC Ltd (0.30) have beta less than 1. It can be seen that majority of securities have beta more than 1.The movement of such securities is high than market movement, hence called aggressive securities. The investors who are willing to take risk invest in such securities because of higher the risk and higher the returns.

CONCLUSION
Stock market prediction is the act of trying to determine the future value of a company stock or other financial instrument traded on an exchange. The successful prediction of a stock's future price could yield significant profit. The efficient-market hypothesis suggests that stock prices reflect all currently available information and any price changes that are not based on newly revealed information thus are inherently unpredictable. Investment plan has a greater chance of success if the investor follows a disciplined approach to each of these factors. The investor's interest is better served by choosing investment products that are different from each other, but where each of them is doing its best to control these factors in its own steady way. In this study, the application of ANFIS machine learning approach was evaluated for Stock market prediction. Several influencing parameters such as adjusted closing price, volume and Net profit have been considered in the ANFIS mode and the results suggest that the ANFIS method can be successfully applied to predict the Stock market movements.