Television and Data Mining
Television - we all have one, and I would venture that many of us watch television shows that can be clustered by similar topics. Television show clustering can be seen simply by noting which channel, or channels, you're watching. Reality tv has been pumped out on almost every channel. Sports networks tend to resemble each other. While I was watching tv last night, I wondered, how do networks gather data, develop new shows based on data, and ensure success of a particular series. I decided to research the topic and stumbled upon an interesting application called the Evolutionary Algorithm (link to article posted below).
The EA (Evolutionary Algorithm), according to the article, has been applied to many real world situations successfully. With this particular approach, researchers are able to look for known and unknown relationships as well as problem solve. Recently, the EA has been applied to television. The EA is able to examine a population that evolves and changes based on certain rules regarding selection, mutation, and recombination. It sounds a lot like biology, but when you think about it in terms of television, it makes a lot of sense. Tv watchers selection shows by preferences which tend to evolve into a variety of shows that or similar, or mutations of a previous selection. These mutations, variety of tv shows, often leads to recombining selections of various shows. Data mining is used to develop initial insight and data trends in order to initiate the algorithm. A sample population is selected. From there the algorithm essentially looks for areas of concentration, or exploitation, among the initial data set. EA then defines parent solutions which based on fitness, or these areas of exploitation. Operators, or samples, are then grouped according to parent functions until the population is grouped into parent functions. The EA is also useful because it does not only correlate population data to its particular parent function with which it fits the best, but it also relates population to each parent function. Networks would then be able to see correlation among channels, shows, etc.
It's obvious that after taking this class, data mining can be applied to any situation. This article is very interesting as it explains a fairly new algorithm that adapts to changing populations.
Very interesting topic. I have ofter wondered when cable/satellite providers would begin to use data mining to provide their customers with "tailor made" television packages. When I think about it, the topics or themes of most of the shows I watch on tv are closely related. It would not be that difficult to provide a "suggested programs" list for the user. Just as the networks mine for popular themes of shows in order to provide what the users want to watch, the cable/satellite providers could do the same for their customers in suggesting programs that may be of interest. This technique you described is utilized by several companies already. If you watch Netflix or listen to Pandora a few times, they begin to create suggested playlists based on what you've viewed previously. Pandora takes it a step further in giving the option of user feedback. Pandora users have the option to "thumbs up/thumbs down" tracks, which helps evolve the algorithm towards the user's taste in music. I hope that cable/satellite providers will soon catch on to this technique.
ReplyDeleteWe hear about the “Netflix” recommendation algorithm that suggests movies and shows within Netflix all the time. However, I have never thought about how data mining could relate to cable television. There is obviously data recorded about who watches what, but I guess it just never really crossed my mind that these statistics like “number of viewers” were used to do anything besides deciding if a show should be canceled or not.
ReplyDeleteI looked into the idea of “Television and Data Mining” a little bit more and found another interesting concept. Data mining is not only used in the cable television industry to figure out what aspects of shows the user is interested in, but it is also used in the part of television watching that most of us hate- commercials. Cable companies rely on advertisers as some of their income. That means that those in the television industry want to save commercials in a world where people are recording the shows they like and are skipping the commercials. A data mining viewer profiling system called the Advertising Delivery System (ADS) has been developed to identify the demographic and behavioral characteristics of viewers based on their viewing patterns. The ADS uses some data mining techniques that we have talked about in class, including what they call “k-nearest neighbor” and the strength of the model. In the discussion of whether or not this methodology will be a hit, it is concluded that “As computer technologies make consumer behavioral data increasingly accessible, it also becomes increasingly important to address the analytical, business, social/privacy, legal, and ethical considerations in data mining for consumer marketing.”
Check out this link to read about this topic in more detail: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CFIQFjAA&url=http%3A%2F%2Fwww.researchgate.net%2Fpublication%2F220427820_Using_data_mining_to_profile_TV_viewers%2Ffile%2F79e415069cdae1bdfb.pdf&ei=vJM4Ua_QEovC9gSPpYGQAg&usg=AFQjCNGGxDKSEdUfoyJ6YPNW1TuFhwUp6A&sig2=uyx7J46X3l-yROAIveSVHw&bvm=bv.43287494,d.eWU