التنبؤ بمبيعات الأغذية باستخدام تقنيات التعلم الآلي

FOOD SALES PREDICTION USING MACHINE LEARNING TECHNIQUES

Abstract

Nowadays, the economy of nations is based on the development of the private sector. One of the most important companies working in this sector is food companies. These companies have suffered from the problem of storage and damage to products due to their expiration date. This thesis proposes building a strategy that predicts the actual need of companies for quantities of products in the presence of other variables. The proposed strategy is based on two main pillars: The first is using three different food datasets with a different correlation between their features, as the first dataset is of high correlation, the second is of a medium correlation, and the third will be a weak correlation. The second is using thirteen Machine Learning algorithms and evaluating their results based on several specific metrics to obtain the best algorithm in terms of accuracy. The obtained results indicate that the best algorithm applied to the first dataset with a high correlation is Gradient Boosting which gave an accuracy (98.65), and the best algorithm applied to the second dataset with a medium correlation is also Gradient Boosting which gave an accuracy (59.29), while the best algorithm applied to the third dataset with a low correlation was Random Forest which gave accuracy (39.39). Based on these results, the proposed model was applied to predict the quantities of Dates that Iraq will produce for the next five years, with the availability of other variables. Where the dataset was of good correlation, the first algorithm was used and gave accuracy (99.51). This thesis sets a limit to the daily food wastage that incurs significant financial losses worldwide.