Automated Agriculture Commodity Price Prediction System with Machine Learning Techniques

Abstract

The intention of this research is to study and design an automated agriculture commodity price prediction system with novel machine learning techniques. Due to the increasing large amounts historical data of agricultural commodity prices and the need of performing accurate prediction of price fluctuations, the solution has largely shifted from statistical methods to machine learning area. However, the selection of proper machine learning techniques for automated agriculture commodity price prediction still has limited consideration. On the other hand, when implementing machine learning techniques, finding a suitable model with optimal parameters for global solution, nonlinearity and avoiding curse of dimensionality are still biggest challenges. In this research, we address these problems by conducting a machine learning strategy study and propose a web-based automated system to predict agriculture commodity price. In the two series experiments, five popular machine learning algorithms, ARIMA, SVR, Prophet, XGBoost and LSTM have been compared with large historical datasets in Malaysia. The results validate the efficiency of the proposed Long Short-Term Memory Model (LSTM) to serve as the prediction engine for the proposed system. Particularly in the long-term experiment testing, the average performance of LSTM with MSE has improved 45.5% while ARIMA has dropped 74.1% and the average MSE of LSTM is 0.304 which outperformed all other four algorithms.

Stacked Ensembles: Boosting Model Performance to New Heights Based on Regression for Forecasting Future Wheat Commodities Prices in Gujarat

“…In comparison to stacked models and multiple regression, model performance is lower at predicting values, as seen by higher MAPE values approximately 11% observed of SVR for 15 days of average prediction. Most of the researchers have focused LSTM (long short-term memory) (2,4) , Neural network (20) , which needs a huge amount of data as compared to the meta-model presented in the current study, and Traditional statistical approaches like ARIMA, SARIMAX (3,5,11) , which required high computing capacity of classical Machine learning models such as autoregressive models, Feature engineering is performed manually and not able to learn more complex data patterns ultimately. Almost all the published research work reported improvement in accuracy with Bi -Direction LSTM.…”

“…An increase in differential order will take more time to run models like ARIMA, SARIMAX. (2)(3)(4)(5)11,20) and Logarithmic Complexity increases with the SARIMAX, Same doesn't apply to LSTM.…”

“…Numerous forecasting strategies have been created in recent years to address the variety and complexity of managerial forecasting problems that are on the rise. The prediction problem has traditionally been solved using linear statistical approaches (such as ARIMA models), but more recently, with the development of machine learning techniques, the focus of the solution has mainly changed from statistical methods to the machine learning domain (4) . Research has been done in recent decades to estimate and predict the price of agricultural commodities.…”

A Machine Learning Framework for Olive Farms Profit Prediction

“…They used climate and soil properties to predict crop production or weather-related events such as rainfall [43,44,101,[114][115][116]. Studies have also focused on financial objectives using only econometric historical observations [99,[117][118][119]. Our work pioneers this approach by combining weather, biophysical, and economic data into an aggregate forecasting framework which is an under-researched area [99].…”

Comparison of Machine Learning Algorithms for Predicting Diamond Prices Based on Exploratory Data Analysis

“…Consequently, an accurate forecast model is crucial for both the investor community and the investment choices of mining businesses. Previous writers have underlined [29] that it is crucial for commodities markets and the global economy to precisely estimate diamond price fluctuations. The price fluctuations of diamonds have been modeled using traditional mathematical and statistical time series prediction techniques.…”