Study of Time Series Forecasting Techniques using NLP, ML and Statistical Approaches.

Abstract

Time series forecasting is a critical task across various domains, including finance, healthcare, and environmental science, where accurate predictions can drive informed decision-making. This study presents a comprehensive evaluation of time series forecasting techniques, encompassing traditional statistical models, classical machine learning methods, and advanced transformer-based architectures. We systematically compare models such as ARIMA and Exponential Smoothing against machine l.earning approaches like Random Forest and Support Vector Regression, as well as deep learning models including LSTM and transformer-based frameworks like Temporal Fusion Transformers and Informer. Evaluation is conducted using standard benchmark datasets, assessing performance based on metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The findings highlight the strengths and limitations of each method, with particular focus on the trade-offs between accuracy, computational efficiency, and scalability. This work aims to guide researchers and practitioners in selecting appropriate forecasting models based on the specific characteristics of their time series data.