The purpose of this project is to analyze and predict my satisfaction towards the restaurants in order to study my food consumption pattern and restaurant preference
Being an indecisive person, I have always had problems with deciding which restaurant I would like to dine in in each meal. My decision making process can take up to hours before coming into conclusion, yet I may not be satisfied with the restaurant I selected. Moreover, as my family of four always have meals together, choosing restaurant is always a burden since we all want to go to a place that will satisfy all of us. This project is purposely designed to evaluate whether I will be satisfied with the restaurant or not based on the restaurant data and my personal opinion in order to reduce the time wasted to deciding restaurants.
This project includes only restaurants that I am familiar with in my regular daily routine. This project does not include restaurant to be chosen on vacation trips.
- What cuisine shows the largest number of satisfied and unsatisfied restaurant?
- What price range shows the largest number of satisfied and unsatisfied restaurant?
- What is the keywords that leads to positive and negative opinion?
- What factors significantly affect my satisfaction towards a restaurant?
Opinion is one factor to affect the satisfaction towards the restaurant. However, opinions are in words and cannot be put into a classification model as the input. To resolve this, using the natural language processing (2-layer-LSTM model), opinions can be classified into sentiment – ‘positive’, ‘neutral’, and ‘negative’ which then can be assigned to value of 1, 0, -1 respectively and can be put into a classification model afterwards.
2-layer-LSTM model is the most suitable for the sentiment prediction because of its advantage in memorizing text patterns. Also, its bidirectional wrapper helps with propagating the input in both forward and backward directions providing higher model accuracy.
In order to show the factors that have large impact on my satisfactory, logistic regression model is used to classify the restaurants into 2 classes – restaurants that will be satisfied and restaurants that will be unsatisfied. Logistic regression method is chosen because its ability to determine the coefficient of each input which is related to the importance of the input on the prediction result. Moreover, the model is used to predict whether I will be satisfied with the restaurant based on the provided data of the restaurant.