Natural Language Processing - Week 1

Sentiment Analysis with Logistic Regression

Lecture: Logistic Regression

Supervised ML & Sentiment Analysis

• repeat until minimize the cost

Vocabulary & Feature Extraction

• sparse희소

Negative and Positive Frequencies

Feature Extraction with Frequencies

Preprocessing

Putting it All Together

Logistic Regression Overview

• dot product greater or equal than zero = positive
• dot product less than zero = negative

• use logistic regression to train a weight factor $\theta$

Logistic Regression: Training

• find $\theta$ parameter that minimize cost function

• gradient descent

Logistic Regression: Testing

• building predictions vector

• validation
  • correct 1
  • wrong 0

Logistic Regression: Cost Function