RecSys Challenge 2015 - Team Budapest

    Features

    • session\_time: unix timestamp of the session.
    • session\_hour: hour of the day @session\_time.
    • session\_hour\_threshold: 2, if session\_hour is between 5 and 18 and 1, if session\_hour is between 3-5 or 18-20, and 0 otherwise.
    • session\_day: day of the week @session\_time.
    • session\_length: length of the session in seconds.
    • session\_length\_diff: difference of session\_length from 1,200 sec.
    • session\_length\_seq: number of clicks in the session.
    • session\_category\_num: number of different categories that have at least one.
    • session\_category\_entropy: entropy of the category occurrence distribution calculated from the session\_category\_num calues in the given session.
    • session\_category\_has\_sale: 1, if the session includes at least one click on an item that is on sale, 0 otherwise.
    • session\_category\_pop: number of clicks on the most frequent category in the given session.
    • item\_session\_dwell\_time: amount of time that the user spent on the given item after the click.
    • item\_session\_dwell\_time\_rel: item\_session\_dwell\_time / session\_length.
    • item\_session\_popularity: number of clicks on the given utem in the session.
    • item\_session\_seq: number of clicks before the given item in the session.
    • item\_session\_seq\_back: number of clicks after the given item in the session.
    • item\_session\_seq\_time: time difference between the timestamp of the click and the beginning of the session.
    • item\_session\_seq\_time\_back: time difference between the timestamp of the click and the end of the session.
    • item\_session\_category: category of the item.
    • item\_session\_category\_pop: number of clicks on the category of the given item in the session.
    • item\_session\_category\_is\_sale: 1, if item\_session\_category is ``sale'', 0 otherwise.
    • item\_session\_category\_merged: special categories are merged into one "other" category.
    • item\_session\_is\_first: 1, if the click is the first one in the session, 0 otherwise.
    • item\_session\_is\_second: 1, if the click is the second one in the session, 0 otherwise.
    • item\_session\_is\_last\_first: 1, if the click is the last one in the session, 0 otherwise.
    • item\_session\_is\_last\_second: 1, if the click is the one before the last in the session, 0 otherwise.
    • item\_popularity: number of clicks on the given item.
    • item\_dwell\_time: sum of time spent by users on the given item.
    • recency\_buy\_first\_backw\_feature: alpha / (\alpha + x), where alpha = 86,400 (1 day), and x=buy\_first\_backw\_feature.
    • recency\_buy\_first\_forw\_feature: alpha/(\alpha + x), where alpha = 86,400 (1 day), and $x=$buy\_first\_forw\_feature.
    • item\_popularity\_buy: number of buy events of the given item.
    • item\_click\_buy\_ratio: item\_popularity\_buy / item\_popularity.
    • item\_price: actual price of the item. Equal to the closest buy event's price when the price was larger than zero.
    • item\_price\_rel: (item\_price - minimum item price ) / (maximum item price - minimum item price).
    • item\_price\_rel\_min: item\_price / (minimum item price).
    • item\_price\_rel\_max: item\_price / (maximum item price).
    • buy\_first\_backw\_feature: time between the actual tiem of the click and the most recent buy event on the given item.
    • buy\_first\_forw\_feature: time between the actual tiem of the click and the next buy event on the given item.
    • buy\_series\_86400\_forw\_feature: number of buy events of the given item in the next one day.
    • buy\_series\_86400\_backw\_feature: number of buy events of the given item in the previous one day.
    • buy\_series\_604800\_forw\_feature: number of buy events of the given item in the next one week.
    • buy\_series\_604800\_backw\_feature: number of buy events of the given item in the previous one week.