Time Series: Analysis and Forecasting using Statistic and Machine Learning Techniques

• Grant number: 24/16370-6
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): December 01, 2024
• Effective date (End): November 30, 2025
• Field of knowledge: Physical Sciences and Mathematics - Mathematics - Applied Mathematics
• Principal Investigator: Thays Aparecida de Abreu Santos
• Grantee: Ana Laura Munarin da Silva
• Host Institution: Faculdade de Engenharia (FEIS). Universidade Estadual Paulista (UNESP). Campus de Ilha Solteira. Ilha Solteira , SP, Brazil

Abstract

A time series is a sequence of data over time, and its modeling is complex due to the presence of both linear and nonlinear patterns. Time series analysis is crucial for identifying trends, seasonality, cycles, and other data characteristics. Time series forecasting is relevant in various fields, such as product demand, electric load, COVID-19, finance, and customer behavior, and can be short, medium, or long-term. Statistical techniques like ARIMA and machine learning methods such as artificial neural networks and SVM are widely used. However, it is essential to analyze the time series before applying any technique to ensure accurate and effective predictions. Many studies fail by not considering this prior analysis, using computationally expensive methods without understanding the historical data patterns. This project emphasizes the importance of analyzing the time series before choosing the most appropriate technique for forecasting.