Television program recommender with interval-based profiles for determining time-varying conditional probabilities

摘要

A method and apparatus for generating television program recommendations based on a viewer's time-windowed viewing history is disclosed. Changes in viewing preferences are automatically identified. The viewing history profile is maintained as a series of viewing history windows. The total length of each individual window and the complete viewing history itself (i.e., the number of smaller viewing windows) can be fixed or changed. In addition, non-critical or non-playable features that can be deleted from the viewer profile without loss of important information are identified. When the corresponding frequency count for the feature fails to exceed the minimum threshold for a predefined period of time, the feature may be deleted from the viewer profile. The disclosed television programming recommender adapts the generated television program recommendations to changes in viewing preferences. The conditional probability of a feature (known classification, for example, the likelihood that programs have characteristics in the classification of the programs watched) is computed as a function of time. For changes in cycle or period in viewing preferences, the disclosed television programming recommender adjusts the condition probability of the property to more accurately reflect one or more similar earlier cycles or periods. For trends or persistent changes in viewing preferences, the TV programming recommender is most likely to reflect the trend (current viewing preferences) best, or the most recent viewing history likely to best estimate the value from the identified trend And generates television program recommendations using the window (s).