Machine learning models for time series forecasting: A case study of coconut price prediction

Abstract

Coconut prices are essential to the economies of many tropical countries, impacting farmers, traders, and policymakers. Accurate prediction models are crucial for informed decision-making and risk management related to price volatility. This paper evaluates and compares the effectiveness of machine learning models and stochastic models in forecasting coconut prices. Using historical data, we examine the performance of various prediction methods, focusing on accuracy, robustness, and practical applicability. The findings highlight the strengths and weaknesses of each approach, providing insights into their effectiveness for coconut price forecasting and offering recommendations for enhancing predictions in the coconut industry. The analysis is centered on the monthly wholesale prices of coconut from Kerala, covering the period from January 1, 1995, to December 31, 2022, as collected from the Indiastat portal (www.indiastat.com). Experimental results show that machine learning models outperform stochastic models across all accuracy metrics.