Predicting Open Restaurants from Yelp Attributes

)

What are the factors that indicate a restaurant is in operation (open)?

This project demonstrates the use of a predictive model to uncover insights on the important factors influencing the model. In this case I used the business attributes for restaurants to predict whether they are in operation (open).

However, the most interesting part was understanding the importance factors for the model and building separate models for restaurants in different geographical areas. Restaurant owners could use the insights to make sure they have the attributes that are desirable for their clientele.

The data used in this report is part of the Round 6 Yelp Dataset Challenge provided for the Data Science Capstone project. The business information is from the yelp_academic_dataset_business.json file.

To predict open, I used the Attributes in the business file including some information on the location (state), hours of operation, and even the average star rating from Yelp.

I also built separate models for state == AZ and state == QC and looked at the importance factors (by mean decrease Gini).

The general random forest model was created with this code:

# Split data set
set.seed(123)

split <- sample.split(rest_df$open, SplitRatio = 0.7)
train <- subset(rest_df, split==TRUE)
test <- subset(rest_df, split==FALSE)

# Build random forest
set.seed(456)
restRF <- randomForest(open~., data=train)
predRF <- predict(restRF, newdata=test)
table(predRF, test$open)