NLP and Sentiment Analysis of Amazon's Baby Products with Logistic Regression in Python

Marcelo Guarido

Dr. Marcelo Guarido will talk about the Logistic Regression principles. He will go through the main ideas of binary classification.

In the Logistic Regression approach, the algorithm sets a separation boundary between classes in a way that maximizes the probabilities for each observation to belong to a specific class. For that, we maximize the log-likelihood (a sum of the log of the probabilities) function and solve its parameters (or weights in machine learning) with a gradient descent solution.

We will test our function on a sentimental analysis project, using reviews of baby products from Amazon to decide if the sentiment is positive or negative. Some natural language processing will be applied to the date prior to the modeling.