Analysis and Classification of Churn in Telecommunication...

Back to Etapa Locală - Sectiunea Claselor 11-12 - Olimpiada Națională de Inteligență Artificială 2026

Competition ended

Archive problem

Competition: Etapa Locală - Sectiunea Claselor 11-12 - Olimpiada Națională de Inteligență Artificială 2026

Problem Description

Customer Churn Analysis and Classification in Telecommunications Services

Context

Customer churn represents the voluntary abandonment of services by customers of a telecommunications operator. Understanding the factors that influence churn is essential for developing retention strategies and optimizing offers and services.

Artificial Intelligence enables complex analysis of customer data for:

identifying customers at high risk of churning
segmenting customers based on behavior and consumption
estimating the impact of different offers and characteristics on churn decisions

To solve this problem, you will have two files available:

train.csv – contains all variables, including Churn, for training models
test.csv – contains the same variables, without the Churn column, and the data in this file is used for generating predictions

Dataset Description

Column	DType	Description
SampleID	int	Unique customer identifier
Age	int	Customer age
Avg Monthly GB Download	int	Monthly average of downloaded data (GB)
Avg Monthly Long Distance Charges	float	Average monthly cost for long distance calls
City	object	Customer's city
CLTV	int	Estimated Customer Lifetime Value
Contract	object	Contract type (monthly, annual, etc.)
Country	object	Customer's country
Customer ID	object	Internal unique customer identifier
Dependents	int	Number of dependents
Device Protection Plan	int	1 if exists, 0 if not
Gender	object	Customer gender
Internet Service	int	Internet service indicator
Internet Type	object	Type of internet connection
Lat Long	object	Latitude and longitude coordinates
Latitude	float	Latitude
Longitude	float	Longitude
Married	int	1 if customer is married, 0 otherwise
Monthly Charge	float	Monthly cost of services
Multiple Lines	int	1 if multiple lines exist, 0 otherwise
Number of Dependents	int	Total number of dependents
Number of Referrals	int	Number of referrals made to other customers
Offer	object	Type of offer received (if any)
Online Backup	int	1 if exists, 0 otherwise
Online Security	int	1 if exists, 0 otherwise
Paperless Billing	int	1 if electronic billing exists
Partner	int	1 if customer has partner, 0 otherwise
Payment Method	object	Bill payment method
Phone Service	int	1 if phone service exists, 0 otherwise
Population	int	City population
Premium Tech Support	int	1 if premium support exists, 0 otherwise
Quarter	object	Quarter in which data is recorded
Referred a Friend	int	1 if customer referred a friend
Satisfaction Score	int	Customer satisfaction score
Senior Citizen	int	1 if customer is senior, 0 otherwise
State	object	Customer's state/province
Streaming Movies	int	1 if movie streaming service exists
Streaming Music	int	1 if music streaming service exists
Streaming TV	int	1 if TV streaming service exists
Tenure in Months	int	Number of months since contract start
Total Charges	float	Total amount paid by customer
Total Extra Data Charges	int	Additional costs for data
Total Long Distance Charges	float	Total amount paid for long distance calls
Total Refunds	float	Total amount refunded to customer
Total Revenue	float	Total revenue generated by customer
Under 30	int	1 if customer is under 30 years old
Unlimited Data	int	1 if plan has unlimited data
Zip Code	int	Customer's zip code
Ping Score	float	Network performance score
Avg Speed	float	Average internet connection speed
TV Type	object	TV receiver type
Link Quality Index	float	Link quality index
Total Short Distance Charges	float	Costs for local calls
Churn	int	1 if customer churned, 0 otherwise

Requirements

For this problem you have to solve 3 tasks.

Task 1 – Simplified Financial Risk Score (20 points)

Calculate for each customer in the test set a simplified financial risk score based on the values of Monthly Charge and Total Extra Data Charges.

Formula:

1FinancialRiskScore = (Monthly Charge > 70) + (Total Extra Data Charges > 10)

Each component adds 1 point to the score if:

Monthly Charge > 70: increased churn risk due to monthly cost
Total Extra Data Charges > 10: additional risk due to extra costs

The final score is between 0 and 2.

Task 2 – Connectivity and Services Score (20 points)

Calculate for each customer in the test set a score that reflects service quality based on the columns Avg Speed, Ping Score and Link Quality Index.

Formula:

Python
1ServiceQualityScore = (Avg Speed < 50) + (Ping Score > 80) + (Link Quality Index < 30)

Each component contributes 1 point to the score.
The final score is between 0 (good experience) and 3 (poor experience).

More specifically, if the first condition (Avg Speed < 50) is met, 1 point is added. Same for the rest.

Task 3 – Churn Classification (60 points)

Build a classification model that predicts Churn (0 = active customer, 1 = customer who churned) based on all information available in train.csv.

You can use all available columns or select the most relevant subset for the model.
Predictions for each customer in test.csv must be 0 or 1.
Answer in the submission file will be 0 or 1 for each customer.

Submission File Format

The submission.csv file must contain:

id - ID taken from test.csv (the SampleID column)
subtaskID
- 1 for Task 1
- 2 for Task 2
- 3 for Task 3
answer - the calculated or predicted value for the customer at the indicated task

Evaluation

Task 1 and Task 2: exact evaluation (comparison with correct answers)
Task 3: for evaluating this task, the AUC metric is used

For Task 3, the score is awarded according to the following rules:

AUC ≥ 0.90 → 60 points
AUC ≤ 0.70 → 0 points
The interval between them: proportional score

Files

Download problem files

Submit Solution

Upload the result file manually. You can also attach source code.100 MB maximum in total

Submission File

Click to upload or drag and drop

CSV, ZIP, etc. (MAX. 100MB)

Source Code File (optional)

Click to upload or drag and drop

Archive, notebook or code file

Work and submit from Google Colab

Open the notebook prepared for this problem. The first cell downloads the data; when you are done, submit your solution for evaluation directly from the notebook.

Colab submissions use the same evaluator and limits. Track the result in My Submissions.

MLCompete can access only the notebook it creates in Drive, not your other files.

Doesn’t open? Use a temporary code

Generate the code here, open the public notebook, and enter it when the first cell asks.

Open the public notebook in Colab

Paste the workspace code only when the notebook asks for it. Never share it with someone else.