Machine learning based restaurant revenue prediction

Abstract

Food industry has a crucial part in enhancing the financial progress of a country. This is very true for metropolitan cities than any small towns of our country. Despite the contribution of food industry to the economy, the revenue prediction of the restaurant has been limited. The agenda of this work is basically to detect the revenue for any upcoming setting of restaurant. There are three types of restaurant which have been encountered. They are inline, food court, and mobile. In our proposed solution, we take into consideration the various features of the datasets for the prediction. The input features were ordered based on their impact on the target attribute which was the restaurant revenue. Various other pre-processing techniques like Principal Component Analysis (PCA), feature selection and label encoding have been used. Without the proper analysis of Kaggle datasets pre-processing cannot be done. Algorithms are then evaluated on the test data after being trained on the training datasets. Random Forest (RF) was found to be the best performing model for revenue prediction when compared to linear regression model. The model accuracy does make a difference before pre-processing and after pre-processing. The accuracy increases after the applied methods of pre-processing.