Commodity Trend Analysis and Prediction Using Machine Learning

This literature survey examines the evolution of prediction methods for agricultural commodityprices, assessing various approaches and highlighting current trends and challenges. Despite theirwidespread use, traditional statistical techniques such as linear regression and ARIMA models oftenstruggle to handle the non-linear characteristics, cyclical patterns, and complex interrelationships typicalof agricultural markets. Machine learning techniques, including Support Vector Regression (SVR),Extreme Learning Machines (ELM), and neural network architectures, particularly Long Short-TermMemory (LSTM), have emerged as promising alternatives to capture these complexities. Combinationmodels, which integrate multiple forecasting methods, have also been explored to enhance predictionaccuracy.Web crawling techniques to gather data from online sources, such as market platforms and newswebsites, expanding the range of potential predictive features. By combining this information withconventional datasets, including past price records and climate statistics a more thorough insight intomarket forces can be achieved. Sentiment analysis derived from social media and news articles asadditional features, reflects the growing influence of public opinion. The review reports the impact ofexternal factors, like global events and climate change, on agricultural prices. Incorporating these factorsinto forecasting models is crucial for accurate predictions and informed policy decisions.