I want to know the Confusion matrix Steps

Question

I've done the Count vectorizer with Cosine similarity. Next, I want the Confusion Matrix to get precision and accuracy

But I don't know how to do it I really appreciate your answers even though they are just steps

let me know if it is wrong / lacking in describe the problem

code Count Vectorizer

    c_vectorizer = CountVectorizer()
    c_vectorized = c_vectorizer.fit_transform(dataset_with_tags.movie_tags)
    c_vectorized_m2m = pd.DataFrame(cosine_similarity(c_vectorized))
    c_vectorized_m2m.head(10)

    c_vectorized_m2m_similarity = c_vectorized_m2m.stack().reset_index()
    c_vectorized_m2m_similarity.columns = ['first_movie', 'second_movie', 'similarity_score']
    c_vectorized_m2m_similarity.head(10)

Answer 1

You seem to be confused about the confusion matrix : it's used when you can compare actual vs. predicted values for a classification problem , thus giving you an absolute truth (TRUE/FALSE) as to whether or not categories were properly identified. Eg how to generate a confusion matrix from the resultswith a classifier .

https://en.wikipedia.org/wiki/Confusion_matrix

Similarity matrices don't categorize , they just provide you with continuous values from 0 to 1 representing how 2 things are similar. There is no classification, thus you cannot use a confusion matrix .

Whether you want to use a similarity matrix (how similar are 2 items) or a classifier (eg whether a movie is a "comedy" or a "drama", movies can have several genres, eg "romantic comedy", so you will need a multi-class classifier), you need some test data to assess the performance of your model :

• Similarity matrix : list of movies which are similar/dissimilar and expect your matrix to return values close to 1/0 respectively
• Classifier : assuming the movie_tags from your dataset are accurate, you can use those to train your classifier, and predict tags for movies which are not in your dataset (you can always use a similarity matrix later on to recommend similar movies based on those predicted tags).