For the telecom dataset the following sub-tasks were conducted:
- I identified the top 10 handsets used by the customers.
- Then, identified the top 3 handset manufacturers
- Next,I identify the top 5 handsets per top 3 handset manufacturer
● Next I Aggregate per user the following information in the column
- number of xDR sessions
- Session duration
- the total download (DL) and upload (UL) data
- the total data volume (in Bytes) during this session for each application
● for this part of the analysis I used the following engagement metrics:
- sessions frequency
- the duration of the session
- the sessions total traffic (download and upload (bytes))
● Based on the above I wrote a python script : - I Aggregate the above metrics per customer id (MSISDN) and report the top 10 customers per engagement metric - Normalize each engagement metric and run a k-means (k=3) to classify customers in three groups of engagement. - Compute the minimum, maximum, average & total non- normalized metrics for each cluster. - Aggregate user total traffic per application and derive the top 10 most engaged users per application - Plot the top 3 most used applications using appropriate charts. - Using k-means clustering algorithm, group users in k engagement clusters based on
Aggregate, per customer, the following information (treat missing & outliers by replacing by the mean or the mode of the corresponding variable):
- Average TCP retransmission
- Average RTT
- Handset type
- Average throughput
Compute & list 10 of the top, bottom and most frequent for the tcp retransmission, Round trip delay and Throughput values in the dataset
- user satisfaction was calculated based on experience and engagement