Consumer Behaviour Analysis in Foodservice Industry

Yang, Poh Yee (2019) Consumer Behaviour Analysis in Foodservice Industry. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Yang Poh Yee_FullText.pdf Restricted to Registered users only Download (3MB)

Abstract

Many companies have realized the importance of business strategies to sustain and maintain a stable income in the market. These business strategies include maintaining the loyalty of existing customers, creating peaks in the customer experience and others. The main objective of this thesis is to extract semantic customers behavioral information through consumer behavior analysis by using churn analysis. The company associated is RunningMan, it is a third-party delivery services that is collaborated with variety of restaurants and fast food stalls. However, due to the increasing number of customers and rapid growth in business size, it is hard for the company to maintain individual relationship with each of their customers. Therefore, they hope to perform some analytics on consumer behavioral based on their customer transactions details, restaurant details and customer profiles. This project focused on using Churn analysis, that is segmented into six processes which are Goal/Objectives, Data Acquisition, Data Pre-processing, Machine Learning, Data Visualization and Data Report. Firstly, literature review and interview were carried out to acquire some domain knowledge on delivery system operations and obtained the datasets from RunningMan. Afterwards, various techniques were used to pre-process the datasets by using like Data Cleaning, Data Integration, Data Transformation and Data Reduction. During data pre-processing, the churn period was defined. Then the cleaned data was fed into various machine learning algorithms and the results were visualized for ease of interpretation. The machine learning algorithms used were Artificial Neural Network (ANN), Support Vector Machine (SVM), AdaBoost Classifier (AB), Gradient Boosting Classifier (GBM), XGboost Classifier (XGB) and Deep Neural Network (DNN). The model was then evaluated using Accuracy, Confusion Matrix and Receiver Operating Characteristic (ROC) Curve. The testing of classifiers was performed with varying features extraction methods and different data sampling size. The outcome of this project includes both the results from the data exploration and the data prediction in which the former refers to the findings and insights regarding the dataset while the latter helps choosing the best machine learning algorithm. However, the accuracy and findings from this analysis maybe undermined due to the small sample size. Since the dataset was collected from February to March 2018, the churn trend can only be tracked within these two months.