I had such a blast making this video! Please bear with the Low Res Zoom recordings lol Watch my newest tutorial! ua-cam.com/video/X1UfLDztdgg/v-deo.html
Thanks a lot for this tutorial. I learned a lot from this.😁 I made a few changes and implemented in a flask app. Would really mean a lot if I could have your opinion on it. 😃😃 My github username is auddy99 and repo name of this project is recoTrax
If I could just have stuff like this that I watch for 4 hours a day, then build for the other 4 hours of a normal workday. I don't have access to people / coworkers with this knowledge which makes it incredibly difficult to immerse myself in it. I think this and building would be a perfect combo! Thanks for the videos!
WOW! Absolutely incredible man, I'm doing my final year project and I'm trying to do some playlist organisation stuff with algorithmic components. This is absolutely what I was looking for, a perfect place to start. I'll be reading deeper into neural networks and such from this video as I want to go for a more complex approach, but I love the stuff man! keep up the awesome work, and thanks for making it so accessible!!
love this my man... my goal is to create a recommender system for my final project and your tutorial has been super clear and helpful in getting me started.
00:02 Building a Spotify recommendation engine from scratch in Python. 01:50 Using a content-based recommendation system to generate recommendations based on summarizing a playlist in Spotify. 05:39 Creating one hot encoded features of the year and a bucketed version of popularity 07:36 Using tf-idf to create features based on song genres. 11:58 The video demonstrates how to build a Spotify recommendation engine using Python. 13:49 The Spotify API provides a Python library with amazing features to easily access and extract live data from Spotify. 17:49 Using recency bias to prioritize songs added more recently in the playlist. 19:38 Generating recommendations based on cosine similarity using a playlist vector and feature set vector. 23:40 Similar recommendations between Spotify and the built recommendation engine 25:14 Closing remarks and feedback request
Thanks for the tip! I’ve started using my iPhone to record and the quality has improved in my newer videos. Check them out and let me know what you think.
Hey! I was very impressed by your vid! Thank you for uploading this for all to see. I literally want to get into machine learning just to do things like this. right now I am going through the Andrew Ng Machine Learning Coursera course. What specific topics do you think I should look at after that in order to understand what was going on in this vid?
I’m glad you enjoyed the video! Andrew Ng’s course is a great place to start, make sure you spend time applying what you’ve learned to your own problems. That’s the best way to make sure you’re absorbing what you’ve learned. As for this video, being proficient in pandas and understanding TFIDF and Cosine Similarity are crucial. They’re pretty common NLP techniques. I provided a brief overview of both in the video!
good stuff! watched till the end, a considerable achievement! how would you ensure that your playlist vector is able to access the entirety of the 'phase space' of song vectors (in the recommendations database). In other words how can you be sure that given infinite unique playlists, all songs eventually get recommended? probably not relevent to the utility of your project, but thats the question that came to me! cheers
This is a great question that isn't often considered! You're right, the current framework wouldn't leverage all of the 'phase space'. To avoid this, I usually incorporate some sort of "reasonable randomness" into my results so that we are able to pressure test a user's interest. Doing this regularly allows you to leverage a larger portion of your catalog while building a more comprehensive user persona. The trick is to identify how much "reasonable randomness" is allowed while maintaining a solid UX.
Hey Ronald, I'm assuming you're talking about the slides? There's a code snippet that allows you to display PNGs in your notebook. Check out the github repo!
I have been running into this error when I try to run the playlist vector: "MemoryError: Unable to allocate 16.4 GiB for an array with shape (13905, 158284) and data type float64" ... curious on if you encountered this/best way to proceed. I appreciate any help!
I was also facing same problem when we transfer it into complete_feature_set but when I run next day it simply run without any error... Sometime kernel creates problem in my case. You can also try by restarting.
As Akshay mentioned, this is typically and issue with your computers memory. What I like to do is restart my kernel and remove any heavy processes while I'm running that code. I hope that helps!
@@akshaysharma6249 is there any chance that you could email me the song dataset earlier uploaded............ as the one currently on kaggle is updated .............i have to submit my college project .... any help would be greatly appreciated
is there any chance that you could email me the song dataset ............ as the one currently on kaggle is updated .............i have to submit my college project .... any help would be greatly appreciated
How were you able to get the access token using the util.prompt_for_user function? I don't really understand the redirect_uri portion and how to get the access token so I can continue.
I was stuck at that part as well. You need to go to "Edit Settings" of the app on your dashboard and set redirect URI there as well. It should be the same as in util.prompt_for_user function. I used 'google.com/' in both places because the localhost address was not working for me.
@@ayushranjan14 Thanks for the help! Once you ran the chunk of code with google did it redirect you to google.com? Currently not redirecting me anywhere and I'm getting prompted to enter the URL I was redirected to. I enter 'www.google.com/' for it both times but still comes out to an error.
@@kevinadventures3965 That happened to me as well. I was using google collab to run the notebook at that time. It started working for me when I switched to the local jupyter environment.
Thanks @Ayush for your responses! You're completely right, this should be run locally which I should have specified. @Kevin, let me know if you are still running into issues.
@@MadhavThaker I'm working with google collabs too and I have this error when connecting to the api it says bad request uri, and it asks me twice for the redirected uri
Can someone help me to find the dataset? I used this link: Kaggle Data: www.kaggle.com/yamaerenay/spotify-dataset-19212020-160k-tracks. But it shows 404 error.
I had such a blast making this video! Please bear with the Low Res Zoom recordings lol
Watch my newest tutorial!
ua-cam.com/video/X1UfLDztdgg/v-deo.html
I am facing problem in id_name its giving me output only this { } any suggestions how I can overcome this problem
Please help me
Thanks a lot for this tutorial. I learned a lot from this.😁 I made a few changes and implemented in a flask app.
Would really mean a lot if I could have your opinion on it. 😃😃
My github username is auddy99 and repo name of this project is recoTrax
@@akshaysharma6249 please can you send me the data please
If I could just have stuff like this that I watch for 4 hours a day, then build for the other 4 hours of a normal workday. I don't have access to people / coworkers with this knowledge which makes it incredibly difficult to immerse myself in it. I think this and building would be a perfect combo! Thanks for the videos!
I'm glad you're finding this content fun and useful. Look out for my next tutorial. I'm planning to tackle text summarization!
Wonderful! Also, I just subscribed and I'm the 1000th subscriber!
Thanks for your support!
WOW! Absolutely incredible man, I'm doing my final year project and I'm trying to do some playlist organisation stuff with algorithmic components. This is absolutely what I was looking for, a perfect place to start. I'll be reading deeper into neural networks and such from this video as I want to go for a more complex approach, but I love the stuff man! keep up the awesome work, and thanks for making it so accessible!!
love this my man... my goal is to create a recommender system for my final project and your tutorial has been super clear and helpful in getting me started.
This is great to hear! Definitely share hour final project when it’s finished
@@MadhavThaker please can you help me with the data, i need it for a project
How does one deploy the model in a cloud platform and then consume in front end app like react. That will be great tutorial. Thanks
00:02 Building a Spotify recommendation engine from scratch in Python.
01:50 Using a content-based recommendation system to generate recommendations based on summarizing a playlist in Spotify.
05:39 Creating one hot encoded features of the year and a bucketed version of popularity
07:36 Using tf-idf to create features based on song genres.
11:58 The video demonstrates how to build a Spotify recommendation engine using Python.
13:49 The Spotify API provides a Python library with amazing features to easily access and extract live data from Spotify.
17:49 Using recency bias to prioritize songs added more recently in the playlist.
19:38 Generating recommendations based on cosine similarity using a playlist vector and feature set vector.
23:40 Similar recommendations between Spotify and the built recommendation engine
25:14 Closing remarks and feedback request
unfortunately, the dataset is not available anymore in kaggle
You sir are an absolute LEGEND!
Thanks! I'm glad you like this content!
You can use apps like droid cam to have your mobile be used as a camera. It would help with the video quality.
Thanks for the tip! I’ve started using my iPhone to record and the quality has improved in my newer videos. Check them out and let me know what you think.
Hey! I was very impressed by your vid! Thank you for uploading this for all to see. I literally want to get into machine learning just to do things like this. right now I am going through the Andrew Ng Machine Learning Coursera course. What specific topics do you think I should look at after that in order to understand what was going on in this vid?
I’m glad you enjoyed the video! Andrew Ng’s course is a great place to start, make sure you spend time applying what you’ve learned to your own problems. That’s the best way to make sure you’re absorbing what you’ve learned.
As for this video, being proficient in pandas and understanding TFIDF and Cosine Similarity are crucial. They’re pretty common NLP techniques. I provided a brief overview of both in the video!
@@MadhavThaker thanks for the reply. I'll refer back to this once I advance further.
Is there a way to make my own recommendations using my own spotify account data?
thank you for making this video its awesome
sir its an excellent video but i have a query, at time 16:04 what is spotify_df we need to give
i can't find the dataset could you please give us the link
good stuff! watched till the end, a considerable achievement!
how would you ensure that your playlist vector is able to access the entirety of the 'phase space' of song vectors (in the recommendations database). In other words how can you be sure that given infinite unique playlists, all songs eventually get recommended?
probably not relevent to the utility of your project, but thats the question that came to me! cheers
This is a great question that isn't often considered!
You're right, the current framework wouldn't leverage all of the 'phase space'. To avoid this, I usually incorporate some sort of "reasonable randomness" into my results so that we are able to pressure test a user's interest. Doing this regularly allows you to leverage a larger portion of your catalog while building a more comprehensive user persona. The trick is to identify how much "reasonable randomness" is allowed while maintaining a solid UX.
@@MadhavThaker thanks, that clears it up actually! :)
We can also personalise the dataset right? or will it be difficult?
Great video! Quick question, what did you use to create the illustrations in your jupyter notebook?
Hey Ronald, I'm assuming you're talking about the slides? There's a code snippet that allows you to display PNGs in your notebook. Check out the github repo!
I wannc write a dissertation about this music recommendations topic. What do you think are the current gaps in this field ?
I am doing a recommender system for a market but I don't know which variables I should take.
I have been running into this error when I try to run the playlist vector: "MemoryError: Unable to allocate 16.4 GiB for an array with shape (13905, 158284) and data type float64" ... curious on if you encountered this/best way to proceed. I appreciate any help!
I was also facing same problem when we transfer it into complete_feature_set but when I run next day it simply run without any error... Sometime kernel creates problem in my case. You can also try by restarting.
As Akshay mentioned, this is typically and issue with your computers memory. What I like to do is restart my kernel and remove any heavy processes while I'm running that code. I hope that helps!
@@akshaysharma6249 is there any chance that you could email me the song dataset earlier uploaded............ as the one currently on kaggle is updated .............i have to submit my college project .... any help would be greatly appreciated
is there any chance that you could email me the song dataset ............
as the one currently on kaggle is updated .............i have to submit my college project .... any help would be greatly appreciated
I also need this dataset. can you provide any other link or send me datasets.
I ran into an error while connecting to Spotify api where it says invalid client after I login what do you think is the issue here
I am facing the same error, Have you found what the problem was? If yes, then pls reply
I solved it by going to dashboard page > Edit Profile and putting the localhost link in the request uri field
The github link isnt working. I cant seem to open the main code file
Can you please provide me the dataset, the kaggle link is showing error 404
How were you able to get the access token using the util.prompt_for_user function? I don't really understand the redirect_uri portion and how to get the access token so I can continue.
I was stuck at that part as well. You need to go to "Edit Settings" of the app on your dashboard and set redirect URI there as well. It should be the same as in util.prompt_for_user function. I used 'google.com/' in both places because the localhost address was not working for me.
@@ayushranjan14 Thanks for the help! Once you ran the chunk of code with google did it redirect you to google.com? Currently not redirecting me anywhere and I'm getting prompted to enter the URL I was redirected to. I enter 'www.google.com/' for it both times but still comes out to an error.
@@kevinadventures3965 That happened to me as well. I was using google collab to run the notebook at that time. It started working for me when I switched to the local jupyter environment.
Thanks @Ayush for your responses! You're completely right, this should be run locally which I should have specified.
@Kevin, let me know if you are still running into issues.
@@MadhavThaker I'm working with google collabs too and I have this error when connecting to the api it says bad request uri, and it asks me twice for the redirected uri
what machine learning algoritnm are you using here
How much u charge for making a video recommendation system for Android app?
Sir can we save this as an model and load it in django
sir can we use it in django also
Guys the dataset has been removed so don't go with it.
Anyone can please share the dataset link.
@gabcode where?
i just farted
INVALID_CLIENT: Invalid client
Can someone help me to find the dataset? I used this link:
Kaggle Data: www.kaggle.com/yamaerenay/spotify-dataset-19212020-160k-tracks. But it shows 404 error.
having the same issues, please if you find a solution tag me