Tutorial 2- Creating Recommendation Systems using Nearest Neighbors

[@mondal1839]

After watching several videos and reading various articles, this one made me understood. Thanks!

[@srishtijaiswal7079]

Can you reply today, I want to know is it item- item colloberative filtering ?

[@mathlearn5476]

Thank you so much Mr.Krish ji, every time when I stuck, your content and explanation will help me lot. The way of your explanation is simply awesome.

[@zagustan]

Hi Sir, This is an amazing video, can you explain more on the way to evaluate the model? thank you

[@Tronyyy]

Khafabi Dza Gustan +1

[@shivalikapatel7222]

how to get recommendations for a particular movie title, instead of a random pick?

[@kumarsaksham5651]

Same question how to get recommendations for a particular movie title, instead of a random pick?

[@srishtijaiswal7079]

Can you reply today, I want to know is it item- item colloberative filtering ?

[@srishtijaiswal7079]

Same question

[@shyam15287]

Love u ji great work ,all the best , continue ur hard work

[@elizavetapudim9677]

Thank you for this tutorial! It also would be cool to hear about more advanced methods, like matrix factorization, autoencoders for recommendation systems, and so on.

[@srishtijaiswal7079]

Can you reply today, I want to know is it item- item colloberative filtering ?

[@elizavetapudim9677]

@@srishtijaiswal7079 yes, they can be applied for item-item cf, as model-based approach

[@lprayaga1]

Awesome videos excellent introduction to the concepts

[@talentzunlimited1398]

Awesome Video ! But many of the questions are still unanswered, it would help everyone.

[@priyankgupta3931]

Sir, how can we evaluate our model in this case? Please reply as I am stuck in the middle of my project.

[@sarabhian2270]

did you get answer on how to evaluate model , I got up to recommending movie but not able to create csv file predicting top 5 movies, can you help?

[@dhirendrajha9667]

Please make one video on deep learning techniques through recommendation System. thank you

[@shaadakhtar5986]

What does distance mean in this one? Is it: for a particular user distance of all movies from the movie at 110th Index. And are you trying to make collaborative or content based

[@srishtijaiswal7079]

Sir, Can you reply today, I want to know is it item- item colloberative filtering ?

[@fangyiyu2744]

Thanks a lot for the tutorial, it's easy to understand and implement, for the recommendation movies at last, what should we do if we want to recommend by the descending order of cosine similarity?

[@srishtijaiswal7079]

Can you reply today, I want to know is it item- item colloberative filtering ?

[@nehajha6273]

Hi Sir, How can we measure the accuracy of this model @krishNaik

[@krishnakumarsaw4917]

How to calculate the accuracy of this algorithm ?

[@animeshkoley6478]

What technique is best for collaborative filtering - cosine similarity or correlation?

[@mohdkashif7295]

can anyone explain how is this collaborative filtering, where did we make use of similar user property used in collaborative filtering, i am getting confused, to me it appear just like content based filtering but in case of correlation we are using cosine similarity.

[@PawanSingh-tz4gy]

Sir, i have seen lots of videos all are similar kind. If you can help with recommendation system where we can use customer features like customer demog(age, education, city etc) and also product features as well.

[@arindammondal9364]

Hi, Thanks for the easy code with a heavy concept. The last output in your code,distances are ascending order. When I increase the neighbors number the more closer distance to 1 is coming. My query is if the distance is closer to 0 is more similar or closer to 1 is more similar(as you told cosine distance more towards 1 is more closer)?

[@garvitgupta13]

What is the difference between KNeighbourClassifier and NearestNeighbour ?

[@questforprogramming]

KNC is for supervised ML and NN is for unsupervised.

[@thomsondcruz]

Hi Krish, How do you suggest handling large datasets through a pandas matrix. For real life data, we get the "Unstacked DataFrame is too big, causing int32 overflow" error

[@DS_AIML]

@Krish,For retail customers,How can we recommend top 3 products to EACH user based on their previous purchase history using KNN or any other better algorithm?

[@sarabhian2270]

did you get the solution for top 3 product recommendation ? if yes then pls guide me by supplying solution

[@ridakhan7934]

If we want to generate recommendations by selecting only one movie rather than random pick. how we can do that? anyone?

[@sarabhian2270]

how to convert this prediction into csv file , I want to predict top 5 movies based on users previous movie, pls help me

[@vibhatripathi3085]

How can you make movie rating prediction system using user-user collaborative filtering and KNN, given rating data (has movieID, UserID, User's rating for the movie) file (.txt) and movie list (.txt) file (has Movie ID, Movie Tiltle, and Year of Release). I think movie list file is just for reference i.e. to find the movie for prediction.

[@brown_bread]

randomly picked 110, is it userId? In the video you are saying, it is a movieId. But after executing pivot table there is no movieId in the table. Am i missing something?

[@nuthalapatirohit7111]

No, its an movie id from the movies data set. in pivot table it shows tops 5 movies. if you search the 110th index manually u can get movie tilte.

[@srijaramarthy5232]

Is it user based collaborative system or item based collaborative system?

[@sitaramchikkala]

its user-user collaborative filtering

[@kamran_desu]

This is item based CF. The rows in this data set are items (movies) that are being used to find their k nearest neighbours. It's similar to finding which items are most correlated. If you transpose the matrix with users as rows, then it will be user based CF.

[@IT_FoodLover]

Sir how I am measure it works properly or not

[@prajwalkhajure3570]

Sir Please make one video on restaurant recommendation System

[@mubasheerkhan822]

wouldn't filling the rating values 0 affect the result in a negative way? because rating 0 is just downgrading the average rating of a particular movie

[@yashkhorania3726]

We are only finding similarity between different movies based on information from different users . We are not rating individual movies, so putting zeroes doesn't downgrade the movie itself. For example: In the video , the 1st recommended movie for Crimson Tide (1995) is shown as Clear and Present Danger(1994), what it means is that the user ratings pattern for both these movies is similar and therefore it got recommended, it doesn't mean either movie is good or bad , just that users rated them similarly. At least that's what I understood.

[@diamondsandpearls2025]

Really great video!

[@Abhimanyukuma_1234]

Sir np.random.choice ke alawa others option movie" title" se function bata do

[@austineferdinand8536]

thanks ,kindly show us the popularity threshold that you concluded to be fifty

[@nandalala7915]

Why can't we use mahanobolis distance instead of euclidean and Manhattan distance. Mahanobolis is a great Indian scientist

[@questforprogramming]

It is not about Indian or foreign person or metric, it is about the performance of the model you build. Metrics are choosen based on cross validation

[@elmerjr.gapuzan9653]

is it possible to include an evaluation metrics?

[@srishtijaiswal7079]

Can you reply today, I want to know is it item- item colloberative filtering ?

[@somalasrestha1283]

sir how to fix "ValueError: Unstacked DataFrame is too big, causing int32 overflow"

[@subhadipghosh8194]

Hii @Somala Srestha You can convert the dtype of int32 to uint8, and then try executing that cell

[@somalasrestha1283]

@@subhadipghosh8194 thank you !

[@SanjeevKumar-if3hk]

How would i get the recommend for a particular movie #krish_naik

[@beizhou2488]

Thank you for providing us these informative and valuable courses for us. I have a question regarding this tutorial. The movie features you were using are actually the ratings given by the users, which should not be considered as the features from my perspective. The movie features should be the attributes of movies, I think. Could you please interpreting this for me to clarify my question? Thank you very much.

[@joeljoseph26]

If its content based filtering then yes, we only use movie features but mostly we use the hybrid model. i.e. (user ratings [collaborative] + movie features [content-based])

[@selvaganapathy265]

rating_popular_movie= rating_with_totalRatingCount.query('totalRatingCount >= @popularity_threshold') Why '@' is used here?

[@bandermaster81]

a variable outside the df

[@srishtijaiswal7079]

Can you reply today, I want to know is it item- item colloberative filtering ?

[@DooDooDarling]

I'm getting an error at this line. 'ValueError: Expression (totalRatingCount) >= (__pd_eval_local_popularity_threshold) has forbidden control characters. '. Does anyone have a solution? I have not been able to get one online.

[@louerleseigneur4532]

Thanks Krish

[@RajasthaniLadka]

whats the motive of csr_matrix(movie_features_df.values)??

[@shubhambaghel219]

A matrix is sparse if many of its coefficients are zero. The interest in sparsity arises because its exploitation can lead to enormous computational savings and because many large matrix problems that occur in practice are sparse.

[@ashishshrestha2275]

is this user-user or item-based recommendation?

[@nuthalapatirohit7111]

its check similarity between two movies, so its an item based recommendation.

[@ilyesbejia6566]

is it javascriipt ?

[@ayushichoudhary6989]

How to dump this values in pickle file please can anyone help it?

[@nuthalapatirohit7111]

training model part convert into pickle file. and import the file where ever you want so that no need to execute whole part

[@rohitsharma-kr9gk]

Sir can you make series video on fitness based recommendation system. and can you tell me how to collect data set of different people and recommend them food and exercise according to that data set. plz reply sir....

[@_ekcup_chai]

What is the data type for title?

[@shindepratibha31]

string: str

[@hamzanaeem4838]

I like the whole series of yours , but this is not good , BAD !!!

[@ajaysaikiranpenumareddy9809]

how to get recommendations for a particular movie title, instead of a random pick?