COMMODITY MARKET TREND: PREDICTIVE ANALYSIS AND CLASSIFICATION USING ADAPTIVE SMOOTH SVM

Abstract

The commodities market, characterized by its inherent volatility and susceptibility to various external factors, presents a significant challenge for traders, investors, and industry stakeholders. In this context, predictive analysis and trend classification are pivotal in informed decision-making. This research explores the application of an innovative machine learning technique, the Adaptive Smooth Support Vector Machine (A-SSVM), as a tool for comprehending and classifying commodity market trends. The adaptive aspect of the SSVM can involve dynamically adjusting the smoothing parameter based on the characteristics of the data or market conditions.The study commences with data collection and preprocessing, encompassing historical market data, price movements, trading volumes, and relevant economic indicators. Kernel functions areused to define the optimized labels for classification from the raw data, while data labeling defines trends as "Upward," "Downward," or "No Significant Change." A-SSVM kernel functions, encompassing linear, polynomial, and radial basis functions, are selected, and the model's configuration is optimized. Predictive capabilities are examined by assessing accuracy, precision, recall, and the F1-score, thereby exposing the model's effectiveness in classifying market trends.