Some feedback: 1) A dropdown (select) option can have a different value than its visual text, so I would use the index from the data loop as the value for every option. Whatever option is then selected, the selected value would match the user index of the data feed. 2) The movies could be in a subarray/subobject. When looping through the data the first time (to create the options) you could re-/move the movie titles into a new subobject for every user, so you wouldn't have to get rid of name and timestamp. less code too. This should be less code overall and hopefully still be understandable for new coders :)
19:06 -> in my head : "oh he made a mistake, he wrote name1 twice" 21:19 -> in my head : "actually, even if you don't, cause you made a mistake earlier. Well, you're going to find out :)"
@@TheCodingTrain Hey man, i love your videos, i clearly understand your explanation but i had a doubt and i want to ask you something, in knn algorithm what if the data looks like this,: user1: { movie1: 5, movie2: 2, movie3: 1, }, user2: { movie1: 1, movie2: 2, movie3: 5, } wouldnt their similarity score be the same even tho their rating for movies are different. And how would recomendation system work in that situation. I would really appreciate if you reply, as i want to learn abt machine learning more.
When you were creating the name lookup table, did you ever consider setting the value of the dropdown menu to be the index of each user and the name of drop-down menu to be to be the human readable name? Then you shouldn't have the problem of doubling the space required for your application to run. Something like `${name}`?
As soon as you remove the nulls you scew your average. Knowing this you might give your nulls your smallest rating -1. This if one person has a null you can still calc without throwing off the average.
your videos help me to keep going through all the programming struggles.Okay here we go. Say you wanted to write a trending system like youtube, how would you approach this ?, if possible, can you please make a trending system. thanks in advance
7 лет назад
May the fourth be with you! (Even though it's may the ninth.)
Hi I try to build in gdevelop. All fine until neighbor now. It should be a farming Programm and in it are different crops who are able to place on a grid. Now my problem is how to know the different neighbors on the grid, who shouldn’t be able to plant near each other. I want to search before I select a new crop from my menu. In example I plant garlic and want to vanish or hide all crops in the menu who can’t be planted around the garlic. So it’s easier to see what exact crops I can plant around each other.
I didn't understand the last example. If Chewbacca is rating every movie 1 or 2 stars and Ewok is rating every movie 4 or 5 then how can they ever be similar. It leads to the conclusion that Chewbacca doesn't like any sci-fi movies but Ewok likes each one of them. The similar shape of the slope shouldn't matter. RIght?
At 26:47 for discussing shortcomings of Euclidean distance , Ewok rates 4s and 5s and Chewbaca rates 1s and 2s resp. , why do you say that even though distance is high their movie tastes might be similar? i did not get that, please could you explain
15:23 Of course! I love your videoes Daniel! (As you can see, we have the same name :) ) Btw 16:57 You could have used forEach, if i remeber correctly?
Hey man, i love your videos, i clearly understand your explanation but i had a doubt and i want to ask you something, in knn algorithm what if the data looks like this,: user1: { movie1: 5, movie2: 2, movie3: 1, }, user2: { movie1: 1, movie2: 2, movie3: 5, } wouldnt their similarity score be the same even tho their rating for movies are different. And how would recomendation system work in that situation. I would really appreciate if you reply, as i want to learn abt machine learning more.
Does anybody know where I can get the original Star Wars data file, the github source code has a different .json file. I'm completely new to working with data so I am unable to translate what he's writing about moives.json so the code works on ratings.json.
Instead of: var titles = Object.keys(ratings1); var i = titles.indexOf('name'); titles.splice(i, 1); var j = titles.indexOf('timestamp'); titles.splice(j, 1); for(...){ ... you may consider: delete ratings1.name; delete ratings1.timestamp; var titles = Object.keys(ratings1); for(...){ ... But this will delete the 'name' & 'timestamp' keys from users object because of the referencing of ratings1 to users. or you can use Array.prototype.filter or just var titles = Object.keys(ratings1); for(...){ var title = titles[i]; if(title == 'name' || title == 'timestamp') continue; ... Just a tip to improve it! :D
Why do you never put the videos that you say will be in the description, into the description? Where's the "What is an associative array" video? Love the vids, but always frustrated when I look for the vid and it's not in the desc.
isn't the Pearson correlation kinda wrong in this case? how someone that rates movies 1-2 has similar taste than I do if I rate those movies 4-5? doesn't he have like the exactly oposite taste?