IMDb Movie Rating Predictor

by Ananya Jain, Manasvi Singh, Pritish Wadhwa and Yash Bhargava, Machine Learning (CSE343/ECE343) under the guidance of Dr. Jainendra Shukla from Indraprastha Institute of Information Technology, Delhi.

Introduction

Is it possible to predict the rating of a movie prior to its release or production? Every year countless movies are made and released worldwide. All these movies are given ratings by viewers throughout the globe.These ratings are combined together to form the IMDb ratings. IMDb rating is the singlemost influential factor in deciding any consumer’s opinion and inherently the success of a movie.

With the machine learning techniques at our disposal, we aim to predict the seemingly unpredictable IMDb rating of any movie before its theatrical release. Successfully predicting IMDb rating is beneficial for both producers (from a financial standpoint) and consumers (from an entertainment standpoint) alike.

Dataset

Links to original dataset: Movie Dataset, Poster Dataset
Link to preprocessed, train, val and test dataset: Complete Dataset

Models

• Regression Models:

  • Linear Regression
  • Lasso Regression
  • Ridge Regression
  • ElasticNet
  • Support Vector Regression

• Classification Models:

  • Logistic Regression
  • Naive Bayes
  • Decision Trees
  • Random forest
  • K-nearest Neighbours
  • Support Vector Machines
  • Artificial Neural Network

Repository Description

• Preprocessing

  Code files for preprocessing data, EDA, feature selection, encoding, train-val-test split and feature scaling

• Regression

  Code files for training, validating, generating graphs and saving regression models

• Classification

  Code files for training, validating, generating graphs and saving classification models

• Reports

  Proposal, Interim Report and Final Report

• Images

  Images used in reports

How to run model ?

Link to saved models: Models

• To load a model
  • .sav model

    filename = 'model_name.sav'
    loaded_model = pickle.load(open(filename, 'rb'))

  • .h5 model

    from keras.models import Sequential
    from keras.layers import Dense, Dropout, Flatten, Input, Convolution2D, MaxPooling2D, Dropouts  
    
    filename = 'model_name.h5'
    loaded_model = Sequential()
    loaded_model.add(Dense(250, activation='relu', input_dim=3561))
    loaded_model.add(Dropout(0.4))
    loaded_model.add(Dense(500, activation='relu'))
    loaded_model.add(Dropout(0.8))
    loaded_model.add(Dense(500, activation='relu'))
    loaded_model.add(Dropout(0.4))
    loaded_model.add(Dense(200, activation='relu'))
    loaded_model.add(Dropout(0.4))
    loaded_model.add(Dense(80, activation='relu'))
    loaded_model.add(Dense(21, activation='softmax'))
    loaded_model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    loaded_model.load_weights(filename)

• To test a model
  • .sav model

    loaded_model.predict(test_X)

  • .h5 model

    loaded_model.predict(test_X)

• To generate graphs
  • Steps to plot graphs are documented in the .ipynb files for respective models

Contact

For further queries feel free to reach out to following contributors.
Ananya Jain ([email protected])
Manasvi Singh ([email protected])
Pritish Wadhwa ([email protected])
Yash Bhargava ([email protected])

Final Report