News All news
29 February, 18:31
We have a new recommendation method (and, accordingly, a recommendation list) "Like recently viewed" The method (and list) "Similar to the ones you liked" we have always had. But it takes as a basis estimates for the entire history of using the Movie Navigator and Imhonet. As a result, I personally (and I suspect many other old users) have this list full of world and Soviet film classics. For the past year, I’ve been watching movies and fiction. And not always really good - often want to see some dynamic nonsense like "Fast and Furious", for which I have forecasts are not so good. And there is no moral strength to watch serious films, and they have been dusting in Deferred for years. So you can assume that I made a new method for myself, although for sure it will be useful to someone else. I like the result so far. . . It works like this: 1. The latest films/series that the user watched (rated) are collected (in an amount of no more than 50 pieces and, if possible, for a period of no more than a year). This takes into account the date of viewing, which is put together with the assessment - i.e. if you note that you watched a long time, these films do not fall into the sample. 2. We select films similar to those on this list. 3. They are sorted by a tricky function that takes into account 4 factors.
But will anyone bother with that?
- Forecast
- The number of films to which they were in similar
- Evaluations of these films
- The degree of proximity to these films
But will anyone bother with that?