Forecasting the Daily Sales of a Franchise

Forecasting the Daily Sales of a Franchise

Sezgi Şener (Istanbul Technical University, Turkey)
DOI: 10.4018/978-1-5225-5137-9.ch006


Historically, restaurant managers used either historical data or simple logical methods to estimate customer numbers or sales volume. These techniques usually consist of an intuitive prediction based on the experience of the manager. However, restaurant sales forecasts are a complex task because they are influenced by numerous factors that can be classified as time, weather conditions, economic factors, and random events. In this case, old techniques may give insufficient results. It is aimed to compare the estimation Simit which is one of the most consumed daily snacks in Turkey sales accuracy of the learning methods and determine the model that provides the highest accuracy and determine the factors affecting the buying behavior of one of the leading Simit chain stores in Turkey in the food sector by using popular machine learning algorithms.
Chapter Preview

Data Analysis

The first step in learning the machine before estimating the data is the Data clearing and visualization process. It is necessary to bring the data that we have found without applying the Machine Learning methods into a processable form. It is necessary to visualize the Basic Effects in the Model. We have received from the company and we interviewed the authorities and we classified the variables as follows. Since our variables are categorical variables, we tried to derive the relationship with the dependent variable mostly with barplots and decision trees.

Table 1.
Important days
1IsNationalHolidayFactor0,1National HolidayImportant Days
Table 2.
Seasonal variations
1IsWeekendFactor0,1Weekend or notSeasonal Variations
2WeekDaysFactor1,2,3,4,5,6,7Weekdays or not
3WeeksDateDate of Weeks
4YearDateDate of Years
5MonthDateDate of Months
6MonthDTFactorOcak-AralıkMonth of the year

Complete Chapter List

Search this Book: