Shared Bike Analysis

The analysis is done by Xinyao Zhang from University of Maryland, College Park. It analyzes and visualizes bike share data from the city of Los Angeles.

with data can be found here.

• Data Preparation
• Most Popular Station
• Average Distance Traveled
• Regular Riders
• Pattern With Season
• Net Change of Bikes
• Category-Passholder type combinations

Data Preparation

1.Data Preview

First, let's take a take a look at the data:

We can find that the starting station ID, starting station latitude and longtitude, starting Lat-Long. All these four columns provide the geography info on which they can support each other when some values are missing.
Then we check the uniqueness of each trip and calculate if there's any value missing from the dataset.

Trip ID                           0
Duration                          0
Start Time                        0
End Time                          0
Starting Station ID              19
Starting Station Latitude        48
Starting Station Longitude       48
Ending Station ID                96
Ending Station Latitude        1051
Ending Station Longitude       1051
Bike ID                          10
Plan Duration                   766
Trip Route Category               0
Passholder Type                   0
Starting Lat-Long             33805
Ending Lat-Long                1051
dtype: int64

2.Filling the Missing Value

By looking at each trip which miss one or more attributes, we choose either to fill in the missing value that can be inferred from other trips, or delete those which are uncertain.

Now the data is clean and can be used!

Which start/stop stations are most popular?

The most popular starting station:

ID:     Number of bikes:
3069    5095
3030    5038
3005    4851
3064    4617
3031    4594
Name: Starting Station ID, dtype: int64

The most popular ending station:

ID:     Number of bikes:
3005    6262
3031    5517
3014    5382
3042    5293
3069    5069
Name: Ending Station ID, dtype: int64

What is the average distance traveled?

Let's calculate numbers of one way and round trip.

One Way       118554
Round Trip     12782
Name: Trip Route Category, dtype: int64

We create a new column 'Distance' based on the distance between starting and ending stations and create a new column 'Speed' based on the distance and duration.

How many riders include bike sharing as a regular part of their commute?

Monthly Pass    80823
Walk-up         40743
Flex Pass        9457
Staff Annual      313
Name: Passholder Type, dtype: int64

Data Visuals: Display or graph 3 metrics or trends from the data set that are interesting to you.

Here we can draw a pie chart shows the Proportion of diff passholder type
this chart can also combined with time series or location

Monthly Pass    80823
Walk-up         40743
Flex Pass        9457
Staff Annual      313
Name: Passholder Type, dtype: int64

How does ridership change with seasons? Types of passes used, trip duration, etc

Trip duration daily pattern

Types of passes used

Monthly Pass

Flex Pass

Walk-up

For Flex pass type, it kept relatively flat all the time.
For Monthly pass and Walk-up, they have some patterns! It seems that when one went down, the other went up

Monthly Pass

Walk-up

There are weekly pattern for the use of Monthly pass and walk-up!

From the plot above we can see that the numbers of passes used experienced periodly drop and rise. On weekdays, the number of monthly pass went up and kept high. At the weekend, the requests decreased quickly and reached its lowest point on Sunday. However, the number of walk-up performed in a opposite way. It went high at weekends and dropped down on weekdays.

The reason for this pattern can be guessed that people who use bike sharing as a regular part of their commute,just like the monthly pass users, will use bikes to work or study more often on weekdays. On the other side, people who don't regularly use sharing bikes, such as the walk-up users, will be more likely to use bikes during their weekends and holidays, when they go outside travelling or sightseeing.

Is there a net change of bikes over the course of a day? If so, when and where should bikes be transported in order to make sure bikes match travel patterns?

There will surely be a net change at each station over the course of day. Because there are so many days and stations to analyze, we think it's more practical to give the recommendation monthly about transportation of bikes based on the net change of each station every month.

Let's draw a heatmap for the bikes net change of each station monthly.

From the heatmap, we can find the max and min values for each month, and then move the bikes from station with max value(which means it has most net bikes coming in) to station with min value(it has most net bikes going out). For example, in 2016-07, the number of bikes in station 3042 increased by 106 and number of bikes in 3068 decreased by 153, so we would recommend that at the end of July, bikes should be transported from 3042 to 3068.

For long-run pattern, we find that numbers of bikes in some stations kept increasing or decreasing at relatively high level.

Increasing station: 3005,3014,3022,3023,3031,3032,3042,3063,3082 Decreasing station: 3007,3024,3027,3028,3029,3030,3052,3055,3068

We recommend transport bikes monthly from increasing stations to decreasing stations according to our analysis. The specific strategy of transportation routes will be decided by the distance and exact change numbers.

What is the breakdown of Trip Route Category-Passholder type combinations? What might make a particular combination more popular?