Web application based on AI/ML techniques for crop price prediction.
This problem can be solved by creating an easily operated web-based application that can be used by farmers all around the country to get reliable price predictions using historical data.
• Using the prior data(at least 10 years) of the commercial crops we can predict a range i.e. upper bound and lower bound of the price using deep learning algorithms such as the N-beats algorithm for univariate time series.
• By including at least one more factor for price prediction we can use bivariate time series. By defining another deep learning model/algorithm for prediction.
• These models aim to beat the naive model i.e. MASE < 1.
• Ensembling the best predicting model on 4 loss functions.
• Crop prices often exhibit complex patterns driven by seasonal cycles, weather changes, and market dynamics. N-BEATS can capture these non-linear and complex patterns effectively without requiring explicit decomposition.