Inventory Managament System

Source Code : Privado

Armazem Paraíba, as well as others retail companies, due to its size and size, has an Inventory Management system, which is continually being improved with new algorithms and business rules in order to maximize costs such as holding cost, transportation cost, shortage cost among others.

Contributions

---

This time i integrate in the project by a creation of an API REST, Able to send the sales forecast of one or more products, along with a suggestion of moving the product over a time window, in order to satisfy that demand and minimize costs.

PreProcessing

---

This was a very challenging part of my contributions to the project, due to some issues with the base, like fact that only recently a database was implemented in this system, until a few months ago these records were made only by hand, so I had to deal with some normalization differences in the data over time, and deal with a range of human errors.

Data Pipeline

---

To ensure that the problems encountered in preprocessing not happen again, I also create a data pipeline to ensure a better environment for future data scientists manipulating these databases.

Sales Problem

---

With the environment ready, I managed a series of methods looking for one that best fit that sales problem for the data available to me. Among the machine learning methods used, the ones that obtained the best results for this problem were XGBoost and Arima.
After Sales forecasting, we calculate a moving suggestion, from the analysis of the company's business rules and the use of optimization algorithms.

API REST

---

For Inventory Management system take these predictions and suggestions, I create an simple API REST, by microframework Flask, with just a few endpoints, to get the data to Sales Problem and post sales updates that go on over time.