FORECASTING STOCK MARKET FOOD INDEX USING COMMODITY FUTURES QUOTES

This article is devoted to testing whether commodity futures can act as tools to predict price fluctuations of commodity indices. We decided to choose WIG-food index and leading commodity futures since they are interconnected. Companies included in our index use commodities in their operations. While we managed to find comparable articles regarding financial market forecasts using different indices, we did not find any opportunities to predict niche indices like WIG-food in our example. Therefore, our article may act as clarification of whether it is possible to use machine learning algorithms combined with basic regression to predict commodity indices price fluctuations. The prediction of stock price movements has long been an intriguing topic of financial research. Particularly prominent ways of predicting a stock price trend include using stocks past performance or using sentiment analysis. However, a certain level of criticism must be applied to these methods. Using past performance creates a certain degree of isolation that leaves out important information carried out by other entities and makes the prediction result vulnerable to local perturbations. Before, we tested whether the relationship between these variables exists to implement machine learning algorithms in price forecasting. Our study consists of Introduction, where were familiarize the reader with the topic; a Methodology, where we describe data collection principles and reasoning for conducting the research, as well as describing machine learning algorithms behind our analysis; Results, where we present the outcomes of our models and a Conclusion. We proposed the use of data collected from different global financial markets with machine learning algorithms to predict stock index movements. In this project, we have attempted to forecast stock price trends using machine learning techniques LSTM. Before we ran basic regression using Keras to derive the reasonability of applying the LSTM model. Regression results showed that commodities included in the model were far from perfection and we managed to achieve 48% accuracy of regression predictive power using commodity futures as features. Therefore, we excluded them from the LSTM model since they turned out to be not credible variables to apply in the machine learning algorithm.