In order to use our Yosnalab shopping services, you need to read carefully about our terms and conditions. If You disagree any part of terms then you cannot access our service.
I agree to pay the rental rate for the period we used the product and in transit and further agrees to promptly return the products at the end of the rental period in the same condition as received.
I will give alert to Yosnalab if my contact information changes.
I also agree to pay for any damages to, or loss of, the rented merchandise occurring during their time of possession or because of loss and or damage of the products. Upon return and inspection if any and all repairs necessary and or accessories missing that were itemized will be charged at our current rates and billed to you.
A full day rental is charged, even for a partial day use. There are absolutely no refunds for early returns.
The products can be used upto 4 weeks only, Cost will be calculated as per day of usage. We are not responsible for the damage of materials or other liability of any kind resulting from the use or malfunction of the equipment. We will not return the initial deposit of money until the product is returned back.
I am responsible for keeping track of my due dates. I understand that any notices sent out by Yosnalab are a courtesy only and failure to receive them does not excuse me from any charges.
A copy of both sides of their valid institution identification card need to be present. Depositing the money can be done only in the form of cash, not through credit/debit cards.
If you have questions or suggestions, please contact us.
Recommender systems algorithms are among the most popular applications of data science today. They are used to predict the "rating" or "preference" that a user would give to an item. Amazon uses recommendation system to suggest products to customers, YouTube uses it to decide which video to play next on autoplay, and Facebook uses it to recommend pages to like and people to follow.
Why: Problem statement
Most of the people were not aware of the good and popular movies. They waste their time on searching it.
How: Solution description
In this challenge, we are going to complete the analysis of Tamil movies in the year of 2017. In particular, we are going to recommend the movies by applying the tools of machine learning such as Simple recommender, Content-Based recommender systems, and Collaborative filtering algorithm.
I collected the list of Tamil movies data from some leading website. The dataset contains Title of the movie, Director of the movie, cast (actors who participated in various roles), Producer of the movie, the year that the movie released, ratings, and the vote counts.
Simple recommenders are basic systems that recommend the top items based on a certain metric or score. In this section, I built a simplified clone of top Tamil movies in the year of 2017.
The following are the steps involved:
Calculate the ratings for every movie.
Sort the movies based on the ratings and output the top results.
One of the most basic metrics is rating. For one, it does not take into consideration the popularity of a movie. Therefore, a movie with a rating of 9 from 10 voters will be considered 'better'.
As the number of voters increases, the rating of a movie regularizes and approaches towards a value that is reflective of the movie's quality. It is more difficult to discern the quality of a movie with extremely few voters.
Taking these shortcomings into consideration, it is necessary that we come up with a weighted rating that takes into account the average rating and the number of votes it has gathered. Such a system will make sure that a movie with a 9 rating from 1000 voters gets a far higher score.
The average rating of a movie on Tamil movie dataset is around 6.29, on a scale of 10. I calculated the average number of votes, received by each movie is 515.
I calculated the metric for each qualified movie. To do this, I define a function, weighted_rating() and define a new feature wr, of which I calculated the value by applying this function to the DataFrame of qualified movies. Finally, I sorted the DataFrame based on the wr feature and the output title, vote count, vote average and weighted rating or score of the top 10 movies.
I calculated the genre for each qualified movie. To do this, I defined a function, build_chart() for the genre, of which I calculated the value by applying this function to the DataFrame of qualified movies.
Content-Based Recommender systems:
I built a system that recommends movies that are similar to a particular movie. More specifically, I computed pairwise similarity scores for all movies based on their plot descriptions and recommend movies based on that similarity weighted ratings. The plot description is available as Overview feature in the dataset.
In its current form, it is not possible to compute the similarity between any two overviews. To do that, I computed the word vectors of each overview, as it will be called from now on. I computed the Term Frequency-Inverse Document Frequency (TF-IDF) vectors for each overview. This will give a matrix where each column represents a word in the overview vocabulary (all the words that appear in at least one document) and each column represents a movie.
I analyzed that 1717 different words were used to describe the 191 movies in the dataset.
I used the cosine similarity to calculate a numeric quantity that denotes the similarity between the two movies. I used the cosine similarity score since it is independent of magnitude and is relatively easy and fast to calculate.
I defined a function that takes in a movie title as an input and outputs a list of the 10 most similar movies. I used reverse mapping of movie titles and DataFrame indices.
Collaborative filtering algorithm:
These systems are extremely similar to the content-based recommendation engine that I built. These systems identify similar items based on how people have rated it in the past. Here I took three columns and found the items that were similar in all the three columns named Director, Genre, and ratings.
By the nature of our system, it is not an easy task to evaluate the performance since there are no right or wrong recommendations. It is just a matter of opinions. This is the RMSE and MAE results between the top 5 movies.
Mean RMSE: 1.9462
Mean MAE: 1.6732
How is it different from competition
I collected the Tamil movie dataset whereas others don't.
Who are your customers
People who are curious about watching various types of movies can use this project. Data Scientists can use this for study and literature survey.