Recommender-System | compared Collaborative Filtering algorithm | Recommender System library

Let us assume that the CountVectorizer gives you a matrix C of shape (N, m) where N = number of restaurants and m = number of features (here the count of the words).

Now since you want to add numerical features, say you have k such features. You can simply compute these features for each movie and concatenate them to the matrix C. So for each movie now you will have (m+k) features. The shape of C will now be (N, m+k). You can use pandas to concatenate.

Now you can simply compute the Cosine Similarity using this matrix and that way you are taking into account the text features as well as the numeric features

However, I would strongly suggest you normalize these values, as some of the numeric features might have larger magnitudes which might lead to poor results. Also instead of the CountVectorizer, TFIDF matrix or even word embeddings might give you better results

I modified your code a bit and got a similar result with mxnet's. Here is the code in colab.

1. model. you missed axis=1 in the summation operation.

Everything is as expected here. 18754 is not the number of training data. This is the number of steps to complete one epoch. The whole training data breaks into a number of groups and each group is called a batch. The default batch_size is 32. This means, your whole training data will be N number of groups where each group contains 32 training data.

So what will be the size of N?

Simple, number of steps (N) = total_training_data/batch_size.

Now you can calculate by yourself.

Btw, this batch is being used because your memory is limited and you can't load the whole training data into your GPU memory. You can change the batch size depending on your memory size.

Embeddings are assumed to be sequential.

The first input of Embedding is the input dimension. So, if the input exceeds the input dimension the value is ignored. Embedding assumes that max value in the input is input dimension -1 (it starts from 0).

https://www.tensorflow.org/api_docs/python/tf/keras/layers/Embedding?hl=ja

As an example, the following code will generate embeddings only for input [4,3] and will skip the input [7, 8] since input dimension is 5.

I think it is more clear to explain it with tensorflow;

You mention "recommended items" so I'll assume you're talking about calculating precision for a recommender engine, i.e. the number of predictions in the top k that are accurate predictions of the user's future interactions.

The objective of a recommender engine is to model future interactions from past interactions. Such a model is trained on a dataset of interactions such that the last interaction is the target and n past interactions are the features.

The precision would therefore be calculated by running the model on a test set where the ground truth (last interaction) was known, and dividing the number of predictions where the ground truth was within the top k predictions by the total number of test items.

Items that the user has not interacted with do not come up because we are training the model on behaviour of other users.

I think one could think about 2 scenarios:

1) if you need very low latency, you can fill up the matrix indeed, i.e. compute all recos for all users, and store it in a key/value backend queried by your app. You can definitely predict multiple users at a time, using the one-hot encoded technique above.

2) predict on-demand by invoking the endpoint directly from the app. This is quite simpler, at the cost of a little latency.

Hope this helps.

Check this sample data:

You should encode the ISBN column as it contains a string using, for example, this snippet

Adding my update as an answer so this can be closed.

I don't think matrix factorization falls into either of the above categories. Source: https://www.youtube.com/watch?v=0sJMMbjjjZM (44:00 minutes in)