presentation_notes_Manupriya.md

Manupriya Sharma:

Introductions and audience interaction, Machine Learning app: Slides 1-2, 8-12

• Introductions - Slide 1

• Fist to five slide for Ryan - Slide 2

• Machine Learning and Interactive App - Slides 8-12

  • Data preparation for machine learning (Slide 9)

    1. Data profiling using pandas_profiling to get an insight how various features affect review score.
    2. Data resampling using undersampling as number of positive reviews are more than negative reviews.
    3. Converting catagorical varibales to numerical variables using label encoder and lambda function.
    4. Converting the target variable to binary using lambda variable

  • Selection of Machine Learning Model (Slide 10)

    1. We selected Random Forest Classification Machine Learning Model out of 5 models (Logistic calssification, KNN, Decision Tree, Random Forest and ANN)
    2. Randome Forest gave us the maximum accuracy of 98%.

  • Feature Selection and Interactive App (Slide 11)

    1. We used random Forest Feature Selection technique to selelct the features that affect the review score the most.
    2. We included zipcode, price, payment value, freight cost, # of photos, Shipping duration, and Shipping delays/early in our app.

  • Recommendations: (Slide 12)

    1. Provide accurate estimations of the delivery times.
    2. Keep the customer updated
    3. Reduce the delivery times
  app_video.mp4