A Machine Learning approach for increasing user engagement in the Fintech Industry
Is it possible to make a user become a regular operator of an application? Make him feel called to use it naturally, as one more task in his daily life? In this thesis, we seek to respond to this not so trivial concern by using Machine Learning as a support tool for the development of two solutio...
Guardado en:
| Autor principal: | |
|---|---|
| Otros Autores: | |
| Formato: | Tesis de maestría acceptedVersion |
| Lenguaje: | Inglés |
| Publicado: |
2023
|
| Materias: | |
| Acceso en línea: | https://repositorio.utdt.edu/handle/20.500.13098/11580 |
| Aporte de: |
| Sumario: | Is it possible to make a user become a regular operator of an application? Make
him feel called to use it naturally, as one more task in his daily life?
In this thesis, we seek to respond to this not so trivial concern by using Machine
Learning as a support tool for the development of two solutions that allow the user
to get more engaged.
As part of a project within a Marketing team of a Fintech company, we seek to
help users go from installing the app to the state defined as ”Habit”. To achieve
this, we take advantage of the available data to develop two Artificial Intelligence
models based on recommendation systems that seek to find the action within the
application that has the greatest chance of being chosen by him.
In the course of this work, some basic concepts (and others not so much) neces-
sary to understand both the business aspects and those related to the more technical
aspect will be introduced.
As a final result, we have developed two models whose objective is to suggest
the next most favorable action for the user, that is, the one that he would not do by
himself but because it was recommended. Always in pursuit of getting the user to
reach the state of Habit. The first of them, a model based on Markovian Processes,
exploits the concept of the Transition Matrix to determine through it the proba-
bility that a person moves from one state (or operation) to another. The second
of the solutions, based on machine learning techniques, seeks to find incremental
suggestions through an Uplift model that determines those actions that are most
likely to generate a positive impact on the user.
With this, we hope to improve the number of users who reach the status of Habit
with respect to current initiatives, thus achieving more committed users and of
greater value to the company, without neglecting their experience or their interests. |
|---|