What is Artificial Intelligence?

With the huge volume of data now available, being able to accurately predict what kinds of information your users want has become mandatory if you want to keep your audience or users.
 
Collaborative Filtering How To Collaborative filtering is a tool that allows for the combination of user input and data sources to predict what your user will need, like or be interested in. It is most useful when dealing with large data sets and pools of users. So far it has been used for mineral exploration, environmental data collection, financial data, commerce and even social networking sites.
 
Collaborative filtering makes automatic predictions about what users will be interested in. It accomplishes this by looking at the past interest of a large pool of users and looking for correlations.

It then assumes that users who liked the same thing in the past will also like the same things in the future. That way the system can use other users preferences to predict the preferences of future users.

This direct connection between similar users is much more accurate and useful than the averages that other systems use. An average tells you what the most common choice is but not what any individual might want. The key to the whole system of collaborative filtering takes two steps. The first step is to find users that share ratings system patterns from past decisions. The second step is to use those shared ratings to predict future ratings for an individual user. The more accurate the results of the first step, the better the results will be from step two.

Collaborative filtering can be used with older information if the information can be sorted correctly. The use of collaborative filtering makes the same information much more valuable. The first company to make use of collaborative filtering was Xerox. They used the system to locate documents using comments from other users. The system had some problems because it only worked if the specific keywords were exact matches. As the system was perfected, it was used with greater success by MIT, Microsoft and Firefly.

A new method of collaborative filtering called active filtering has been developed in recent years and is increasing in popularity. Active filtering uses a peer to peer filtering system that makes the internet much more accurate than before. Instead of just measuring and recording user action it allows users to rate items and publish it on the web for everyone to see.

These ratings can then be used to predict the preferences of future users.It can rule out totally irrelevant search results by comparing them to the highest rated results of other users with similar interests.

There are some potential problems with collaborative filtering due to the fact that it's not a completely passive system. It needs the input of users instead of just passively harvesting information. Another potential problem arises when you're dealing with smaller sample sizes. A few very biased reviews could skew the entire rating system. The future of collaborative filtering will certainly include safeguards that will eliminate these problems because its benefits too great for companies to not want to take advantage of it.

One company that uses collaborative filtering successfully is Netflix, the online DVD movie rental company. Customers pay a monthly subscription to Netflix which allows them rent as many movies as they want each month. This means that customers are constantly visiting the site to look up new movies. Netflix has a star rating system that allows users to rate how well they liked the movies they've already seen. Netflix then uses collaborative filtering to make recommendations on other films they might like.

It does this by comparing their past preferences to that of other users and then finding others with similarly rated movies. It then looks for other highly ranked movies by those users and then recommends them to the primary user. Netflix also shows the average star rating for all users for a particular film, but also shows a predicted rating for an individual using the same collaborative filtering technique.

The collaborative filtering method works well for Netflix for several reasons. First of all, they have a captive audience who is constantly returning to the site and on whom they have collected a considerable amount of data. Second, the users have a clear benefit from participating in the collaborative filtering since it generates recommendations for them that lead to a better movie watching experience. Perhaps most importantly, the system works because Netflix makes it so easy for customers to use. One click of the mouse is all it takes to rate a film.