(The) prediction of industry stock index using artificial neural network

Abstract

During the past 30 years, many studies have witnessed the predictability of stock price in market. With the great breakthroughs known as AI (Artificial Intelligence), frontiers who search for the consolidation between stock market and AI, make a continuous efforts to accomplish their ultimate aims. Given that traditional methods are placed under unrealistic assumption, unquestionably AI technologies such as Neural Network may lead current stock market to more efficient and systematic as well as profitable one with great sudden spurts. The supplementary indicator ISI(Industry Stock Index) is used to analyze the current state of particular industry in the stock market. With a view to play a role of benchmark, more accurate ISI is desirable. As to Korea stock market undergoing a differentiation on the basis of industry sectors, the prediction of ISI functions as a good benchmark for stock selection and an alternatives for portfolio construction. Setting our research aim to the prediction for the directions of next month ISI, we performed conceptual grouping a factors influencing ISI as macroeconomic, industry, Intermarket, and derived factors by economic aspect. we extracted significant input variables for the developed model from these factors: simple time series model, Intermarket, and composite model for each industry. With different input variables, each model was experimented by the three layered feedforward Neural Network. Equipped with global strategies which vary in progress of learning, we obtained the results of best models for Construction and Banking respectively. For CISI, the best network yielded the 80% accuracy(hit ratio), while BISI didn*t get a good performance with no significant difference between models by (66%). As an initial step for performance evaluation, the results show the NN outperform Multiple Regression for both accuracy(80% vs 63%) and learning capability (correlation coefficient is 0.94 vs 0.48) for construction. In the second ...