Data-driven analysis of climate impact on tomato and apple prices using machine learning.

Area of Science:

• Agricultural Economics
• Data Science
• Environmental Science

Background:

• Limited research exists on applying machine learning for agricultural product price prediction.
• Environmental factors are hypothesized to indirectly influence fruit production and prices through crop growth.
• Understanding the temporal dynamics between environmental changes and price fluctuations is crucial for agricultural markets.

Purpose of the Study:

• To assess the accuracy of predicting tomato and apple prices using environmental data.
• To quantify the impact of individual environmental factors on fruit prices.
• To explore the role of time lags in the relationship between environmental variables and fruit prices.

Main Methods:

• Utilized machine learning techniques, specifically Long Short-Term Memory (LSTM) networks, for price prediction.
• Modeled the data-driven relationship between environmental factors and fruit prices, incorporating variable time lags.
• Employed Shapley Additive Explanations (SHAP) to determine the importance of each environmental factor.

Main Results:

• The study successfully predicted fruit prices by incorporating environmental data and identified optimal time lags.
• Including time lags significantly improved the accuracy of price predictions.
• SHAP analysis provided insights into the influence of specific environmental factors on fruit prices.

Conclusions:

• Machine learning models, particularly LSTM, can effectively predict agricultural product prices by accounting for environmental variables and time lags.
• Identifying and integrating appropriate time lags enhances prediction accuracy.
• This data-driven approach offers valuable decision-making support for agriculture, especially in the context of climate change.