I have a neighborhood polygons shapefiles which contains information about population and a point shapefile of all pharmacies in the city. Now I want to create the map where I can see the best locations for my new pharmacy depending on population and existing pharmacies locations. I want locations to be symbolized "best to low" with different shades of colors on the map.
Which steps and tools I should use to solve my problem? I have a ArcGIS 10.4
The network analyst has lots of appropriate tools
An overview of the Network Analysis toolset—Help | ArcGIS Desktop (for Pro but the same for arcmap)
Things like weighted overlays, or even simple overlays.
Of course things like your basic descriptive statistics of your data aka choropleth mapping should keep you busy for a while. Start simple, focus on the end-goal once you have an appreciation of your data.
There's no one tool that will solve this problem for you. As Dan says, start simple and get more complex as you refine what you want to do. For example, you have two main problems to solve: 1.) calculate population density (Point Density, or possibly Kernel Density), and 2.) calculate distance from existing pharmacies (Euclidean Distance). Combine with Map Algebra. Move on to more complex tools once you've decided the simple tools don't do what you want.
OK. So, in my case I need to calculate polygons centroids and to create Kernel Density map using population field? After that I need to calculate distances between pharmacies using Euclidian Distance tool and after all I'll have two rasters. Which operation with Map Algebra I need to use with these rasters in order to get what I need?
Oh, if you've got population polygons, then just convert them to raster. The idea is that you'll get two rasters, each with a "score" (population and distance) and combine them in some way so you have a single index of high to low. Perhaps you'll scale the population values to 0-1 range, and distance to 0-1 range. Add them together and you'll have an index from 0-2. But, play around with your data and figure out what works for you - I haven't put that much thought into it.