Can you forecast if a corporate banking client will be active in the following week?
The objective of this challenge is to create a machine learning algorithm to determine if a user performs a specific action (such as making a purchase) in a 6-hour period over the course of a day, based on previous event data for seven weeks.
There are 4 200 customers, across five countries over a period of seven weeks. The training set is 7 weeks of data, and the last day of the final week will be the test set, where the target is whether a client performed event 14 during that day at 6-hour intervals.
This challenge is also unique in that the data is provided in the same way it would be encoded for machine learning - testing participants' data analysis and reasoning skills from the outset, having to discover the logical relationships before building models.
Variable Definitions
eventdatetime: event date and time in “Y-m-d H:M:S” format
event: name of the event (will be “signup”, “purchase”, etc)
userid: the ID of the specific user logged in
dayofweek: day of week the transaction was made on, you can use df['dayofweek'] = df['eventdatetime'].dt.day_name() to get the day of the week
useremaildomain: user’s email provider (“gmail.com”, “yahoo.com”, etc)
userrole: used to indicate user level – regular employee through to financial director etc
companyprofileid: company ID
country: jurisdiction of the user (South Africa, Botswana, etc)

Can you predict a customer’s income based on their transaction history?
The transaction data is over a period of 14 months, running from the start of July 2021 to the end of August 2022. There are 5144 customers and 46926 unique accounts, with some customers having more than one account. The number of transactions per customer range from 1 to just over 2 000 for the 14-month period.
The train set contains 3600 customers along with their declared net income and the test data contains the remaining 1544 customers with the declared income excluded.
Along with the train and test files is the transaction history for each customer which details the various transactions each customer engaged in over the entire period of recording. There are also data files containing demographic information on each customer as well as files describing some of the categorical variables related to the customers.
How to use Colab on Zindi
How to mount a drive on Colab
The objective of this challenge is to create a machine learning solution to determine a customer’s income based on their transaction history over 14 months. Your model can make use of all the data provided with the target being the declared net income from the train file.
Variable Definitions
employment_status.csv - contains descriptions of the various categories of customer employment status
income_group.csv - describes how customers are grouped based on their income
Test.csv - contains customer identifiers as well as record dates for customers in the test set
SampleSubmission.csv - a sample of the test file with zero predictions on the declared net income
Train.csv - contains customer identifiers as well as their declared net income(the target) used to train your model
customer.csv - contains demographic information on the customers that can be used as predictors in your model
transactions.csv - customer transaction history for the past 14 months