Loading...
Theses
Browse
1 results
Search Results
ThesisItem Open Access Forecasting and multi-criteria decision making for trading and investment in stock market using soft computing techniques(G. B. Pant University of Agriculture and Technology, Pantnagar, 2022-07) Bisht, Kiran; Arun KumarDuring the last decades the globalization of economies, intensifying competition among firms, financial institutions and organizations as well as the rapid economic, social and technological changes, have led to an increasing uncertainty and instability in the financial market. Within this context, the importance of making efficient financial decisions while handling complexities has increased. Soft computing covers a wide range of techniques offering tolerance to the presence of uncertainty and imprecision, making them ideal for modeling financial decisions. In light of the foregoing facts, the present research is centered to propose financial decision making models based on soft computing techniques which include fuzzy sets, neural networks, and evolutionary algorithms. In this study, seven models are introduced. Model [1] presents multi-objective financial trading systems based on deep reinforcement learning. The systems have been designed to give signals (buy, hold or sell) in the live market in order to get maximum return with minimum risk. Reward-based deep reinforcement learning architectures have been used to simulate the models for gaining better results. Model [2] proposes a financial time series forecasting method for stock selection. Fuzzy time series forecasting procedure using fuzzy c-means clustering and two deep learning architectures (SVM and MLP) has been developed for forecasting daily stock prices. Portfolio construction based on stock prices and ACO algorithm for portfolio optimization is derived. Model [3] considers the concept of diversification addressed by modern portfolio theory and suggests a hybrid MCDM technique for diverse stock selection. Neutrosophic base-criteria method has been devised for criteria weight assessment. The concept of correlation coefficient between assets with PROMETHEE partial ranking are merged for finding realistic relationship between the stocks to construct a diversified and profitable portfolio. PSO algorithm for optimizing Portfolio’s Sharpe ratio with rank constraint has been derived. Model [4] provides an effective portfolio construction method based on sector analysis. Dempster-Shafer theory and Granger causality test have been employed for indentifying strong and diverse sector of economy. Construction of portfolio by picking leading stocks of strong and diverse sectors and its optimization using deep recurrent neural network is presented. Model [5] offers a stock selection method integrating the concerns of a novice investor and a stock market specialist. An integrated framework unifying fuzzy delphi method, fuzzy base-criterion method and Dempster-Shafer theory has been developed for assessing important criteria, their weights and reliable ranking of stocks. Portfolio optimization is done using LSTM embedded deep recurrent neural network. Model [6] introduces fuzzy TOPSIS and evidence theory based intraday stock selection procedure. To counter the ambiguity and conflict in intraday data belief divergence measure has been employed for credible ranking of intraday stocks one day before trading. Model [7] presents an MCDM method hybridizing MEREC and CoCoSo method through some specific modifications to their main structures in context to its application in ranking stocks. The method has been developed in neutrosophic environment to cope up ambiguity and inference of decision making data. Parabolic measure has been used as performance measure in MEREC method to reduce its complexity. All the present models have been implemented on real data of Indian stock market (NSE and BSE) and detailed analysis have been done to verify their practicality.
