Comparison between cities

Overall the order of average rating: San francisco > Albuquerque > Detroit
The order of average number of reviews: San francisco > Albuquerque > Detroit
Since San-francisco is the largest city among the three and its residents come from all over the world. It is not surprising SanFrancisco has the largest number of reviews given a large population base. Also, restaurants in international cities are severing people from all over the world. They are also facing tougher competition given the number of their peers. Therefore restaurant in Sanfrancisco tend to provide better food than restaurants in other two cities, which explain higher rating. Also, from the view of customer, it is also possible that customers in San Francisco are more tolerate than customers from other two cities.

Price vs Number of Reviews

We also discover that in three cities, the restaurant with most number of reviews always falls in the range of low price. It is natural that the restaurant with lowest price attract most customer given the supply rule.

Rating and Price

In three cities, there are also a trend that, the higher the price for the restaurant, the rating for that restuarant are more likely falling into a high level. Since expensive restaurant usually offer high quality food, better service and a friendly environment.

Introduction

From the Yelp website, we find the relatedness of each restaurant. It is known that Yelp recommends those related restaurant for similarity between them. For instance, restaurants may sell same kinds of food or are close to each other. Therefore we want to visualize such similarity between those restaurants which are related. In the network graph, each node is a restaurant and edges are connections. Here we color the nodes by their tags( after cleaning). We are hoping we can see some grouping on the nodes of the network graph.

Note: Since it is possible for each restaurant to have multiple tags, we add boolean attributes to each restaurant( such as whether a restaurant has tag "Chinese" or not). Therefore, if we click on tag in the legend, the colored nodes are restaurants with that tag, while the gray one are restaurants does not have that tag.

Since github Markdown does not support Iframe, the images is not interactive for now,however it is linked to the original webpage at plotly. The interactive plots are showed there. If you want to see whether restaurants with tag "Chinese" are grouped well, click all the tags on the legend that is not chinese. Then it will only show you the restaurants attributes with "Chinese" or not

Similar to the analysis above, tags that are grouped well

Chinese Restraurants, Japanese and Kocrean restraurants, American restraurants, SouthEast Asian Restaurants, Indian Restaurants, SouthAmerican food

Less satisfying

Those are tags that does not group well from the plots:
Acohol, Dessert, Europe.

The reason why restraurants with those tags does not group well are:
Firstly there are other information like location involves in the connetion plots;
Secondly, the number of restraurants in those plots are relatively larger than the restraurants with tags like Chinese, Japanese and Kocrean, making them harder to group.

Distribution

The number of restaurants with tags:
American > Acohol, Dessert, Europe ,SouthAmerican> Chinese, japaneseKorean, Indian > others

Chinese,Alcohol,Kocreal and Japanese,South American,South East Asian,Indian,Europe, Dessert

Restaurants with those tags does not grouped well in San-francisco

American

Number of American Restraurants is the largest in San-francisco, and they share most connections. However, the grouping of American restaurants may simply due to the large sample size.

Distribution

The number of restranrants with tags: American > Dessert, Alcohol, Europe > Chinese, Japanese and Kocrean, SouthEast Asian, South American > Indian, others

Above all, restaurants in San-francisco does not group well as the other two cities. We noticed that the number of restaurants in SF is much larger than Detroit and Alburquque. Therefore, those restaurants may scattered spatially, making them hard to share same collection. Also, in san-franciso, it is more like for one restaurant to obtain multiple tags( given san-francisco is an international city). Therefore those tags lost specifility here.