30. Text: Recap

### Recap

Types of Recommendations

In this lesson, you worked with the MovieTweetings data to apply each of the three methods of recommendations:

  1. Knowledge Based Recommendations
  2. Collaborative Filtering Based Recommendations
  3. Content Based Recommendations

Within Collaborative Filtering, there are two main branches:

  1. Model Based Collaborative Filtering
  2. Neighborhood Based Collaborative Filtering

In this lesson, you implemented Neighborhood Based Collaborative Filtering. In the next lesson, you will implement Model Based Collaborative Filtering.

Similarity Metrics

In order to implement Neighborhood Based Collaborative Filtering, you were introduced to and applied a few techniques to assess how similar or distant two users were from one another:

  1. Pearson's correlation coefficient
  2. Spearman's correlation coefficient
  3. Kendall's Tau
  4. Euclidean Distance
  5. Manhattan Distance

Types of Ratings

We took a quick look at different types of ratings:

  1. Did the user interact with an item or not.
  2. Did the user like an item or not.
  3. More granular scales 1-7, 1-10, etc.

It is important to understand what the data might be used for, and what type of granularity might be important for a particular case. One of the main considerations is whether you want to have neutrality available, in which case an odd number of possible values in your scale will provide a value in the middle. Another common question is, how many levels do you really need to understand how much a user likes a particular product? Again, this is largely up to individual preference and specific use cases.

Business Cases For Recommendations

Finally, you looked at the four ideas needed for businesses to implement successful recommendations to drive revenue, which include:

  1. Relevance
  2. Novelty
  3. Serendipity
  4. Increased Diversity

At the end of this lesson, you will have gained a ton of skills to build upon or to start creating your own recommendations in practice.

Next Lesson

In the upcoming lesson, we will take a closer look at model based collaborative filtering, different methods for dealing with the cold start problem, and how to assess how well our model is performing. Then as a final touch, you will have the opportunity to deploy your recommendations to the web!