| Title |
Gross merchandise value prediction for a buyer in e-commerce |
| Translation of Title |
Bendroji pirkėjo prekių vertės prognozė e-prekyboje. |
| Authors |
Žilinskienė, Monika |
| Full Text |
|
| Pages |
42 |
| Keywords [eng] |
GMV, CLTV, e-commerce, XGBoost, FFNN, multi-output FFNN |
| Abstract [eng] |
In order to make informed marketing decisions on customers' acquisition, retention strategies and amount worth to spend on them, business aim to know Gross Merchandise Value customer will create in a given future. In this thesis, besides already researched methods, the application of two- stage model using gradient boosted decision tree and two-stage feed forward neural network, where first stage is classification model to predict if customer will churn in a given future period and second stage is predicting GMV for those whom churn prediction is negative. Further, we propose a multi-output feed forward neural network for combined results: classification for churn and regression for GMV. Finally, models were created on real life customer-to-customer marketplace dataset and evaluated in between using the mean absolute error (MAE) and root-mean-squared-error (RMSE) metrics. |
| Dissertation Institution |
Vilniaus universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2022 |
