Dataset for Hybrid Deep Learning Model for Wheat Price Forecasting

This dataset comprises 38\u202f,019 daily transaction records for various wheat products collected from a Turkish commodities exchange between 1\u202fJune\u202f2022 and 4\u202fMay\u202f2023. Each row contains the transaction date, product class and name, estimated quantity, and unit price, along with 18 quality\u2011related attributes such as moisture content, hectolitre weight, protein percentage, insect damage, foreign material, weed contamination, broken or shrivelled grains, and other defect metrics. The resulting multivariate time series enables research on wheat price forecasting, quality assessment, and anomaly detection. The data are provided in CSV format and were used in our study \u201cA\u202fHybrid Deep Learning Model for Wheat Price Forecasting: LSTM\u2013Autoencoder Ensemble Approach\u201d to train and evaluate ensemble models combining long short\u2011term memory (LSTM) networks and autoencoders.