"The students of the 2nd year of Mechatronics Engineering were challenged to develop digital solutions - using Machine Learning and IoT - to reach consumers of the Huggies and Plenitud brands. The Winners were: 1st place - Team 7 (Bruno Albuquerque/ Felipe Pardini / Luiz Henrique / Paulo Fernando)... "
Kimberly. A multiplatform artificial intelligence that learns from customer's responses and classifies them using Machine Learning. Our Team developed it using Python, Java for Android to create an app, and IBM Watson to create a chatbot.
It is also able to respond with pre-recorded voice messages, which gives it charisma.
This project is part of our annual challenge, 'NEXT'. Kimberly-Clark was judged us by our pitch and demonstration.
I was responsible for the text-classifier, for the Virtual Assistant lines and recording and for the pitch and demonstration of our project to Kimberly Clark managers.
I used Python 2.7 and the following statistics models for the text classification:
Ada Boost Classifier
One Vs. Rest
One Vs. One
The code had an 83% accuracy with base data, and it reached 67% accuracy only 130 lines of real data, the percentage increases proportionally to the number of real data lines. Still, it will reach a maximum of 98% accuracy.