Data

The participants are provided with several datasets:

General Data

• clients.csv: general info about clients;
• products.csv: general info about stock items;
• purchases.csv: clients’ purchase history prior to communication.

Task-Specific Data

• uplift_train.csv: a subset of clients for training. The column treatment_flg indicates if there was a communication. The column target shows if there was a purchase afterward;
• uplift_test.csv: test subset to estimate uplift value;
• uplift_sample_submission.csv: example submission file with predictions. Your submission should be in the same format.

Submissions format

For building the model, participants are provided with a training set of clients uplift_train.csv (see above) with information on performed communication (treatment_flg) and purchasing (target).

For each client in the test sample uplift_test.csv, evaluate the effectiveness of communication (uplift). Submit the result as a CSV file with two columns: client_id and uplift, for example:

client_id,uplift
008fb49e3a,0.1149912020897228
0095340acc,0.8353208872466903
015c0b4d79,0.3085840952650095
...
ff70c360ad,0.0809048695228205
ff86a1311b,0.4815832858531034
ffcccc2cc4,0.10523347182011245

Please note that the uplift’s absolute values are not important, as they are used only to rank clients.

Baseline solution is provided.

Making sense of uplift

There are three possible outcomes when we decide whether or not we should send a text to the client:

• +1: the client made a purchase after we reached out, and probably she or he would not make a purchase if we didn’t reach out
• 0: regardless of us reaching out, the client would make a purchase (or would not make a purchase, which is equally significant in this case)
• -1: the client did not make the purchase after reaching out, but if there was no communication, she or he would.