Can you tell me that the meaning of the items order & repetition in "Impressions.csv"?

[recsyschallenge]

Can you tell me that the meaning of the items order & repetition in "Impressions.csv"?

[mogami95]

And i also want to know if you shifted all timestamps in "Impressions.csv" (year-week) & "interactions.csv" (unix timestamp) with the same way? Are the timestamps in these 2 files still consistent?

[dkohlsdorf]

Yes they are shifted in the same way. Also there is no ordere assumed in the impressions for a given row

[douglashiwo]

how about the meaning of repetition？ I mean for some impression ，posting ID like 1386585 can repeat several times and more ，can we treat it as the number of times that the previous RS had recommended for a single item(posting？)
Looking forward to your reply! thank you

[BrightFeather]

How should we use this data? I mean, is the impression a benchmark for examining the correctness of our result?

[dkohlsdorf]

Well a job can be recommended multiple times to a user. One example of what you can do with an impression is to distinguish between things shown by the recommender or searched by the user directly. However, we don't say you have to use the impression.

Daniel

[mogami95]

@dkohlsdorf Thanks for your explanation. And my last question is: Whether the "impression" was the results of your baseline algorithm(https://github.com/recsyschallenge/2016/blob/master/Baselines.md), or not?

[dkohlsdorf]

Nope the baseline is not related to the algorithm run on the platform. The algorithm on the platform has way more components and filters. So relating the baseline algorithm to the impressions might not be advisable.