Forecasting the daily number of customers in each restaurant

Project Details

Term: 

Fall 2016

Students: 

Edison Lee, Celia Chen, Sehyeon Jeong, Guan-Jie Chen, Web Yuan

University: 

NTHU

Presentation: 

Report: 

Business Problem

We are going to let manager of each restaurant know how busy they will be tomorrow by this business forecasting. The forecasted value would be used as a mental preparation of manager.

Data

The data is provided from iCHEF. It was taking the form of invoice which includes information of timestamp, items, and so on. And then, we make it into 2-column data (demonstrated in the right side of this paragraph) with preprocessing. We are measuring the daily number of people to forecast future number of people dine in
the restaurant. Every restaurant has weekly tendency. Weekend has more customer than weekday. Some of restaurants show a slight decrease of people. The others have no specific decrease or increase.

Forecasting Solution: The manager will receive the result of this forecast one day ahead. They would have mental preparation for tomorrow, and prepare next day job allocation with our forecasted value.
Forecasting Limitation: The lack of data amount harms the accuracy of future forecasting. Because of it, if we provide to manager “Under-forecasted value”, we might let manager mishandle the job allocation.

Recommendation

To complete this project, we tried to choose the different number of days forecasting. We found the performance of 2-day and 3-day ahead are not bad. Based on this result, we recommend not to choose only 1-day ahead forecasting so that the manager can get the forecast earlier. In addition, we highly recommend that iCHEF provides data on daily basis for the suitability of forecasting.

Application Area: