1st place in the SENAI/SC - CTAI Analytics Hackathon, in partnership with the company AMBEV, which took place on August 2nd and 3rd, 2019.

The case was the “ideal assortment for points of sale” and the target audience were the company's salespeople and their service at points of sale (POS) that were already company customers. The objectives were to increase the sales volume of products (SKUs) that were already purchased by the POS and explore the opportunity to offer other SKUs to complement and maximize the brand's mix at the POS.

The solution implemented was a web application in Flask, where for each POS the system generates a suggested product mix with estimated sales volume based on similar POSs. For this, an exploratory analysis of about 5 million data from the database provided by Ambev was carried out, data processing and use of machine learning models, including KNN to perform the search for similarity between POSs and Random Forest to estimate the volume of sales for the new POS.

2nd place in the SENAI/SC - CTAI BIG DATA Hackathon, in partnership with the company Confederação Nacional das Indústrias - CNI, which took place on November 8th and 9th, 2019.

The case presented by CNI was the “identification of the profile of employees on leave from the industry” and the target audience was SESI employees who identify problems, investigate the causes and offer solutions for partner industries. The challenge, then, was to identify the causes of absence, prevent and reduce absenteeism for health reasons of employees of Brazilian industries.

The solution implemented was an online platform for consultations and analysis of social indicators that allow the presentation of diagnoses of the impact of an employee's profile on his or her absence due to health reasons. For this, RAIS data from the years 2008 to 2011 were used, approximately 256 million data. These were loaded into 4 Stages in the hive through PySpark and partitioned into 200 blocks to enable data manipulation. Dimension tables of informative variables were also worked on, with their descriptions. Based on these data, the Naive Bayes machine learning model was used to estimate the probability of absence through the profile of the employee.