Forecasting Stock Price Movements to Pivot Point-Based Support/Resistance Levels Using Artificial Intelligence

Abstract

Stock Trading, especially day trading, is a challenging task. Pivot Point Analysis, being a leading indicator, is an important tool for day trading, but the literature review suggests that there is a research gap in using artificial intelligence-based systems to forecast pivot levels which the price is expected to reach in future. This study aims to outline a system designed using Long Short-Term Memory (LSTM) to map the price movements along with multiple technical indicators taken as inputs to predict the next level, among the pivot levels, that the price is expected to reach in future.
The study proposes a method to map the pivot lines which change daily for day trading to a single output column so that it is possible for an artificial intelligence-based model to efficiently learn from it. The study also outlines the model architecture for designing such a system using Long Short-Term Memory networks and after conducting 44 different experiments, suggests the optimal configuration parameters for the proposed model. The research also proposes different metrics to evaluate the proposed model. The study conducts 6 more experiments to indicate that the proposed model works across different types of stock assets like indices and individual stocks, and across different timeframes for day, swing or positional trading. Finally, the research formulates an illustrative trading strategy to use the proposed model efficiently in real-world scenarios.