- Exploratory Data Analysis (EDA) - During the initial stages of my project, I delved into Exploratory Data Analysis (EDA) to make sense of the dataset I had downloaded. This involved digging into the data, cleaning it up, and visualizing it to gain insights into its structure and characteristics. I removed any features that didn't seem relevant, aiming to streamline the dataset for further analysis.
- Approach - When it came to building a recommendation system, I looked up on google and came across content based filtering system and collaborative filtering system. I decided to take a content-based filtering approach. I figured that utilizing the descriptions of movies and TV series would be a easiest way to make recommendations. I employed cosine similarity, a technique that calculates the cosine of the angle between vectors representing the descriptions. This helped me measure the similarity between different titles based on their content, giving me a foundation for the recommendation engine.
- Initial Attempt with Flask: - I initially tried to implement the recommendation system using Flask, a Python web framework that combines the python backend with html, css and javascript frontend. While Flask offered great flexibility, I realized that it demanded a deeper understanding of backend development, which I am still learning. The complexity was a bit overwhelming for my current skill level and was hitting on multiple errors and "Error 404" while trying to launch the app, so I felt that I needed a simpler solution.
- Transition to Streamlit - Streamlit proved to be incredibly user-friendly and intuitive, allowing me to focus more on the frontend and user experience rather than getting bogged down in backend intricacies. I quickly set up text input boxes and buttons for users to interact with and display the recommendations seamlessly.
- Install streamlit: Since the app works on the base of streamlit, this package is necessary to execute the program
pip install streamlitTo run this project on local machine follow these steps
- Clone or download the repository as zip file and extracrt it.
- Install necessary dependancies
- Download the dataset in the same destination where the files has been extracted.
- Open a terminal, navigate to the project directory containing your script, and run the Streamlit app.
streamlit run app.py
- Limited to Description: One limitation of the current system is that it generates recommendations solely based on the description of movies and TV series, without considering their genres. Need to work on the model to take into account both the description and the genre. This would result in more refined and context-aware recommendations, catering to a wider range of user preferences.
- Dataset Dependency: The recommender currently relies on the titles present in the dataset. Any input title not found in the dataset yields no recommendations. To overcome this limitation and offer a seamless experience, I need to train the model on a new and a bigger dataset. This expansion would ensure that users receive recommendations for a broader array of titles, both popular and obscure.
- Enhanced User Interface: While the current user interface serves its purpose, there's immense potential for improvement. Future work could focus on making the UI more interactive and intuitive. Incorporating features like user profiles, ratings, and user history could significantly enhance user engagement and personalization, elevating the overall user experience.
- Introduction To Recommender Systems- 1: Content-Based Filtering And Collaborative Filtering
- Build A Machine Learning Web App From Scratch
- Streamlit Documentation
- Custom Streamlit Background Image/Color Gradient through CSS
- How to Hide Menu & Customize The Footer in Streamlit Apps
I greatly value feedback as an essential tool for my growth. If you have any feedback to share, please share it at [email protected]. Your insights are immensely appreciated and contribute to my continuous improvement.
