Chapter 10

How the Data is gathered for recommender system?

1. Explicit rating that users provide. ratings after watching the movie 2. Implicit search engine queries and purchase histories. listening to a song 3. Other source of knowledge about the users/items

How are Association Rules are created?

1. Searching data for frequent if-then patterns. 2. Using the criteria support 3. Identifying the most important relationships.

How do you measure User Similarity using Pearson's Correlation?

A measure of how strong relationship between two variables. Degree of linearity can be determined using Pearson Correlation. It determines whether linear component of association between two continuous variables. It has no assumptions about linearity. Pearson correlation is not used to determine the strength. -1 highly similar , + 1 not similar After getting similarity predict

What is Association Rule mining?

Association rule mining uses machine learning models to analyze data for patterns or co-occurrence in a database. it follows association rules which is to identify frequent if then associations. 1. Each transaction is considered to be a list of items. 2. Association rule find all rules that correlate the presence of one set of items with that of another set of items 3. It identifies frequent patterns 4. It is mostly commonly used for market basket analysis.

How do you select the value of K? and what is its range?

Choice of K nearest neighbors for the neighborhood formation results in a tradeoff. Particularities of user's preferences can be blunted and a large k impacts precision. The optimum k depends on data characteristics. K is set in the range of values from 10 to 100.

What is Collaborative Filtering?

Collaborative Filtering systems make recommendations based on historic preference of the users. Such as items clicked watched purchases rated.. User item matrix can be used to present these preferences. It uses filter based on 1. Item-based nearest neighbor 2. User-based nearest neighbor

What is User-Based nearest neighbor?

User-Based filter recommends items by finding users similar to active user.

What is Cosine Similarity?

It produces better results in item-item filtering. Ratings are seen as vectors in n-dimensional space. Similarity is calculated based on the angle between the vectors.

What is adjusted Cosine Similarity?

It takes average user ratings into account. Transforms the original ratings.

What is Item-Based nearest neighbor?

Item based nearest neighbor is preferred over user-based approach due to dynamic nature of users. Items based approached is easy to scale and can be computed offline and served without constant re-training. Can be best implemented through KNN model.

What is Recommender system and how it is useful?

Many e-commerce and retail companies implement recommender system on their websites to leverage the power of data and to boost sales. Goal of recommender system is to predict users interest and recommend products that user may be interested in. Recommender system are one of the most successfully and wide spread application of ML in business. It is a information filtering technique that provides user with recommendations for which they might be interested in.

What is Market Basket Analysis?

Market Basket Analysis is a modelling technique based upon the theory that if you buy a certain group of items, you are likely to buy another group of items. Itemset : Set of items that a customer buys and Market Basket Analysis is used find relationship between purchases.

What is purpose of recommender systems?

Prediction Perspective: Predicts to what degree a user likes an item. Most popular evaluation scenario in research. Interaction Perspective: Gives user a good feeling. IT educates users about the product domain. Conversion Perspective: Looks for commercial situations Increase hits and clicks throughs Checks lookers and booker rates Retrieval perspective: Reduces search costs by providing correct proposals Recommendation perspective: Identifies items from the long tail where user don't know about the existence

How Recommender systems are useful in day to day life?

Recommendation systems act as a solution for day to day choices.

How recommender systems make recommendations to users?

Recommends items that are most popular among all the users. It divides the users in to multiple segments based on their preferences and recommend items to them based on the segment they belong to.

How are similarities measured?

Similarity is measured based on two algorithms. Cosine and adjusted cosine.

What are performance measures for Association rules?

Support : It indicates how frequently the item appears in the data. provides fraction of transaction that contain X and Y Confidence: Indicates the number of times the if-then statements are found true. It indicates how often X and Y occur together, given the no. of times X occurs. Lift : Compare the actual confidence with the expected confidence. It indicates the strength of a rule over the random co-occurrence of X and Y.

What is Apriori Algorithm?

most commonly used algorithm to discover association rules Candidate Items sets for Apriori algorithm are created using large item set of the previous pass. This large item set is joined with itself to general all item sets with a size larger by one. Each generated item set with a subset that is not largest is deleted. The remaining item sets are the candidates. 1. It uses frequent item set to generate association rules. 2. Support value of frequent item sets is greater than the threshold value. The algorithm reduces the number of candidates being considered by only exploring the item sets whose support count is greater than the minimum support count.