The Application of Neural Network and Logistics Regression Models on Predicting Customer Satisfaction in a Student-Operated Restaurant

A student-operated restaurant has to balance the achievement of its objectives as a profit generator and as a learning centre. This unique characteristic distinguishes a student-operated restaurant from other types of restaurant. This study aims to build a model to predict overall customer satisfaction in a student-operated restaurant. The input variables consist of 32 dining service attributes, which are derived from DINESERV factors. Data was collected using a close-ended questionnaire and was distributed using a convenience random sampling approach. A neural network model and a logistic regression model were built to predict overall customer satisfaction. The result shows that the best neural network model built in this study was the MLP neural network model with two hidden layers. The correct classification rate of this model was 80.65% and 69.81% for the training and testing data set. The top three important attributes that influence overall customer satisfaction are customer satisfaction toward service, responsive service and excellent service. In addition, the best logistic regression built in this study was a stepwise approach. This model had a correct classification rate at 73.39% and 69.17% for training and testing data set. The result of logistic regression shows that two significant dining attributes that influence overall customer satisfaction are customer satisfaction with service quality and food quality. Based on the correct classification rate, this study concludes that a neural network model has a better performance to predict overall customer satisfaction than a logistic regression model. However, a neural network model may not be the best model to determine the most significant input variable toward an output variable since it cannot be proven using a statistic method.