Potentially good areas to look for dinner in LA

Los Angeles County Department of Public Health publishes the data about restaurant and market inspections, including addresses of the facilities, their coordinates, scores, and grades of inspections, cities where they are located, the status of the program (active or inactive), and kinds of violations. In this project, I used data from 2018 to June of 2020 and only chose restaurants, excluding markets, and filtered out inactive programs. I was left with 26 203 observations. On their own, dots of restaurant locations create a pretty pattern that follows the main streets network but it’s only possible to get some information when you look closer.

That is why I decided to divide the county into 1 square kilometer hexagon cells and calculated the number of restaurants in each of them. The result is shown on the map below. Most of the area is dark — 50% of the cells contain less than 7 restaurants — with separate bright isles. The distribution of the number of restaurants per cell is far from uniform — there is a high peak on the smaller values with a very long tail going up to number 297. This outlier is located in Downtown.

Probably, it is evident for anyone who visits Los Angeles and especially lives here, but this visualization shows especially clearly where city life is active. These areas are shown in bright yellow color which not only stands out but also in my intention reminds the inviting interior lights. These lights in their largest concentration lead one’s view from Downtown to Hollywood and through Beverly Hills to Santa Monica. Kind of a corridor of lively places with busy sidewalks and tourists there, isn’t it?

Locations of restaurants in the city are interesting but not a very difficult thing to guess. The data from the Department of Public Health lets us see another less obvious thing — their objective “quality” in the form of scores from health inspections. Scores are connected with grades in form of letters A, B, C. In the dataset, scores lower than 80 are classified as C, 80–90 points is B, and 90+ points is A. Good news is that 93% of restaurants have grade A! Because of this unbalanced quality, I worked with scores because they are more precise and show more variation.

For each restaurant, I took mean scores from all the inspections held there, and then took mean scores of all the restaurants located in each cell.

A final measure that I wanted to get from my data is where there are many restaurants, and where they are good. This metric doesn’t measure how sophisticated and tasty the dishes are or how stylish the interior is. Doing that is worth developing a whole own methodology and might be very subjective.

The measure on the next map consists of two variables described above— the number of restaurants in a cell and their mean score in it. Because of the very different distributions, simply overlapping two measures would give a skew of the final one to the places with lots of restaurants. Bigger counts completely absorb any quality values, and that is why I needed to balance them. I decided to take three steps:
1) classify both measurements into 5 classes to bring them into comparable form and also to “dim” extreme values
2) give more weight to quality by multiplying the class number by 2
3) add these two and classify data again.

The result keeps that lively corridor in the western part of the city while highlighting those eastern cities with good scores. There also seem to be some responsible restorators in Northridge and Chatsworth because two of the 12 cells with scores higher than 99 are located there.

The project was done with ArcGIS Pro and the data was prepared with R Studio.