Posts by Collection

portfolio

Algorithms to predict Employee Attrition

It is well known that human capital is one of the main assets of any company, if not the most important one. For this reason, it is wise to lower the attrition rate as much as possible. To do this, we must first know who are the employees that are quitting and what are the variables which might be explaining this behavior. With this purpose in mind, we wanted to create an algorithm that could predict employee attrition. On this occasion, we worked with a fictional open dataset created by IBM data scientists.

Movies Recommendation System: GroupLens Dataset

Users usually consume a new product or service based on recommendations made by other users. This is clearly seen when deciding whether to watch or not to watch a movie. Companies such as Netflix use recommendation algorithms to predict how many stars a user will give a specific movie. Unfortunately, their data is not publicly available. However, the GroupLens research lab generated a dataset with over 10 million ratings for over 10,000 movies by more than 69,000 users. We used this dataset to create a movie recommendation algorithm

publications

Dishonesty in the presence of moral emotions

Published in , 2020

Acts of dishonesty, such as corruption, are considered one of the main country-level problems in Peru. The main objective of this research was to evaluate interventions designed to reduce dishonesty, as well as to explore the psychological mechanisms that could be involved. Through the conduction of two experiments, we wanted to evaluate whether including the pictures of identifiable victims could reduce dishonesty. The first study was conducted in person with the general public (n = 40) in a shopping center; the second, online with university students (n = 100) from a private university in Lima. To assess dishonesty, participants performed the “die task” 20 times in both studies. The degree of dishonesty was measured in aggregate, at the group level. The results of study 1 and 2 show findings that differ from each other. On the one hand, the results of Study 1 show that participants assigned to one of the experimental conditions lied to benefit a single victim, at the expense of their personal financial gains. On the other hand, the results of Study 2 show that participants assigned to the same experimental condition lied to benefit themselves, but lied less than participants assigned to the control condition. The presence of moral emotions possibly generated by the experimental conditions are discussed. Future research will include more precise psychological measurements that allow better identification of the mechanisms of action of the designed interventions.

Download here

talks

teaching

Applied Econometrics

Graduate taught, Pontificia Universidad Católica del Perú, 2022

The course focused on learning the main causal inference tecniques, when performing an RCT is not possible.

Big Data vs Good Data

Graduate taught, Pontificia Universidad Católica del Perú, 2022

The course was divided in two sections. In the first part we explored sampling methods, whereas in the second, we dived into parallel computing.

Deep Learning

Graduate taught, Pontificia Universidad Católica del Perú, 2022

The course focused on some applications of deep learning algorithms; the programming language used was Python.

Machine Learning

Graduate taught, Pontificia Universidad Católica del Perú, 2022

The course covered the main supervised and unsupervised algorithms in Machine Learning; the programming language used was R.

Natural Language Processing

Graduate taught, Pontificia Universidad Católica del Perú, 2022

The course equipped students with practical tools needed to perform natural language processing tasks; the programming language used was Python.