A Neural Network Approach to Predict Stock Prices of ESG Companies Using Technical Indicators

Abstract

This paper explores the application of neural networks, specifically the Multilayer Perceptron (MLP), for short-term stock price prediction of companies aligned with Environmental, Social, and Governance (ESG) principles. The study analyzes data from 100 ESG-compliant firms across 16 sectors, covering June 2024 to June 2025. Technical indicators such as MACD, EMA, Signal Line, and Volume were employed as input variables. Data preparation was conducted using Microsoft Excel, while model training and validation were performed in IBM SPSS. The findings reveal that the MLP model achieved high accuracy, with most prediction errors within ±1 of actual stock prices. Variable importance analysis highlighted the closing price and momentum indicators as dominant factors influencing predictions. Scatter plot validation confirmed a strong correlation between predicted and actual values. This research demonstrates that combining ESG-focused company selection with AI-driven modeling provides reliable tools for sustainable investing. Future improvements may include integrating ESG scores, sentiment analysis, and macroeconomic indicators.

Keywords