Predicting movie ratings was never this easy,
ORRR
Should I say it never felt this bot-like 😎
- NumPy – For arrays
- Pandas – For DataFrames, reading csv, etc.
- BeautifulSoup – For Scraping web pages
- Requests – For getting requests from URLs
- Multi-Threading – For Running multiple functions simultaneously in different threads
- Matplotlib – For Plotting graphs for analysis purpose
- Seaborn – Based on matplotlib, advanced graphs
- Scikit-Learn – For various models, data encoding, metrics, etc
- Major Cast
- Director
- Duration
- Year
- Budget
- Gross
- Votes by Users
- Certification
Some things we can infer from the above graph are:
- As the time changed, the movie budget and gross increased as well.
- Also, as the time changed, the number of voters also increased because of more access to internet.
- Duration had no effect with years, and are randomly set, but also major movie with higher time duration were released in later part of the graph
- Movie with rating higher than 9.0 weren’t affected by time.
Rating of the movies. Rating will be out of 100, and on the scale of 5. Rating is in the Range of 70 (7.0) - 95 (9.5). Reason for such scale: Data has been taken only from Top 1000 Movies list on IMDB and after web scraping - cleaning data - we get only 661 Valid Movie data.
Dataset made with love by me!! Made with the help of web scraping. Code to Web Scrapper: IMDB top 1000 Scrapper Cleaning of the data is in the file Trial.ipynb , Linear_Regression.ipynb and some piece in Web_Scrapper.ipynb The code has been distributed all over the notebooks due to rush, hope you understand.
We will be encoding the following columns:
- Year
- Director
- Major Cast
- Certification
To encode these as per requirement, we will be using Label Encoding Once we encode the data, we export it into temp_data.csv for further use.
We split the data by general means, by using train_test_split, having a ratio of 0.90:0.10 (90% : 10%), 1500 random states and true shuffling.
Before building the model, we will be using Standard Scaler on each model and also use Pipelining to train multiple models at once. We use the following models:
- Logistic Regression
- Gradient Boosting
- Random Forest
- Ridge Classifier
Accuracy varied as per random splitting of data. Mainly 65% Greater for all models at major times, which is great for self-generated dataset. Sometimes it even crosses 80%, which is like a sweet cream over the cake.
Really need someone to scrape some more data out of IMDB, rather than me just getting only valid 60% from 1000. We can have some more, can't we ?? Have some good ideas ? Have the knowledge to contribue to this project ? Or maybe you found a bug, glitch ? Or maybe you can help to clean this code ? Or maybe anything you want to share ? Create an Issue request or Pull Request and we can carry the talk over there !! I am always open to suggestions :)
In future, we can improve the accuracy to higher rates by adding more data into dataset, use different models, and even include neural network if possible. A different encoding technique and addition of steps like PCA can also result in better output. We can also analyse the data into more different types. We might add more columns of different data types in future to improve the feasibility of the project.
Distributed under the License of MIT, which provides permission to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software. Check LICENSE file for more info.
Made till here ? I would like to thank you for reading this whole(possibly) README to reach this endpoint. If you found this software helpful in any case, make sure to star it. (┬┬﹏┬┬) (┬┬﹏┬┬)