The purpose of this project is to demonstrate the data visualization techniques in matplotlib. As usual in this course, the weekly challenge builds upon the code that was developed throughout the modules for the subject matter in question.
The analysis begins with 2 csv files. The first contains data of city names, number of drivers, and type of city. The second contains city names, date & time, fare, and a ride ID. These files are read into pandas dataframes and inspected to determine if there are any missing values by using the isnull function.
Once the datasets are inspected and explored, they are combined with the pandas merge function.
After the dataframes are merged, we can subset the data into additional dataframes and use the groupby function to gather counts and means that will be used as the x & y axes for scatter & bubble plots.
Assembling all of the bubble charts together provides an appealing approach to review multiple sets of data at once.
Fig. 1:
From here we begin to look at summary statistics and develop box and whisker charts for ride counts, fares, and driver counts. These charts are helpful in visualizing central tendencies, spread, and outliers in one image.
Fig. 2:
Fig. 3:
Fig. 4:
Pie charts are developed to show percentages of total fares, number of rides, and number of drivers by city type.
Fig. 5:
Fig. 6:
Fig. 7:
A summary dataframe is built, formatted, and displayed. Finally, a multi-line chart is developed by organizing the data using pandas pivot_table and resample functions to create weekly bins of data. A great deal of formatting was done to create an attractive chart.
As expected, there are fewer rides and fewer drivers in the rural and suburban areas as compared to the urban areas. Average fares increase as we move away from the urban areas since the rides in suburban and rural areas tend to cover greater distances.
Fig. 8:
The multi-line chart shows the same phenomenon. The Urban fares are highest, and the fares decrease as we move into the Suburbs and Rural areas. The trend over time appears to be stable and flat. However, this data is only the first 4 months of the year. It would be interesting to see how the data looks over the remainder of the year.
Fig. 9:
- Programming: One of the largest programming challenges for me was researching and discovering the pandas pivot_table function. This was used to reconfigure the dataframe to set up the weekly bins by city type that would be used for plotting the multi-line chart.
- Programming: Another challenge was figuring out that I did not need to "unpack" the Date column from the index in order to change its datatype to datetime. This was accomplished with the code below.
-
Data analysis: Once the datasets are merged, the number of drivers for each city is duplicated throughout the dataframe (i.e. each instance of a city will show the number of drivers). To overcome this, we must use the city data dataframe to calculate the total number of drivers per city type. If the programmer does not recognize this, there will be a large error in the results.
-
Graphing, etc: A lot of time was spent formatting the multi-line chart. My largest challenge in this area was uncovering how to change the format of the timestamp x-axis labels to simply show the month abbreviation. I discovered matplotlib.dates and, after reading some of the documentation, was able to use the DateFormatter to get the desired format.
-
Description of Approach
As shown in the boxplots of Driver Counts, there is a lot of variation in the number of Urban drivers (see Fig. 4 above). Which cities have the most drivers? How do ride counts and total fares stack up to the number of drivers by city? -
Technical Steps
- Create bar chart for driver count by city
- Format and sort to show city with highest driver count to lowest
- Look in the Urban cities dataframe, groupby city, sum fares
- Create bar chart for total fares by city
- Look in the Urban cities dataframe, groupby city, count rides
- Create bar chart for total rides by city
- Combine bar charts and format accordingly (investigate multi-axes charts)
- Depending on outcome, use average fare instead of total fare
-
Description of Approach
In order to optimize the number of drivers, we should look at the pattern over time that the number of drivers follow. Do the number of drivers follow the same trend every week? every day? Unfortunately, the data provided only shows a static number of drivers per city. If we are to ascertain the staffing level of drivers over time, we would need to have data on the number of drivers at the time of each ride. -
Technical Steps
- Request data tracking
- Receive csv's, read into a dataframe
- Inspect for missing data and explore data provided
- Merge datasets if necessary
- Groupby cities
- Set index to datetime
- Create pivot table to show data over time by city
- Resample data to create weekly and then daily bins
- Plot data on multi-line chart