Time Series: Predict the Web Traffic
-
PriceHardman 9 years ago This question is directed at those three or so people who got 50/50 on the leaderboard: It looks like your submissions consisted of printing out hard-coded literals. Did you
(a) End up making use of leakage from the sample data,
(b) find the hidden dataset out on the internet somewhere, or
(c) actually manage to predict the 30 values with 100% accuracy?
My guess is (b), but I'm sure there are those (myself included) who are curious.
Also, just as an aside @PRASHANTB1984: Great contest! Thanks!
-
stdiff 9 years ago After the deadline of the contest we can get test cases at the "Problem" page. (Click "More" in the right column.)
The deadline was wrongly set at the beginning of the 7th day. (See the first thread.) Therefore I guess that the test cases could be seen until the deadline was fixed. (I did not find them actually, so this is just a guess.)
-
PriceHardman 9 years ago Ahh that would do it. Thanks.
-
-
PRASHANTB1984 9 years ago Thanks PriceHardman :) BTW - all those 50 point guys hard coded the answer. Either a data leak from the source, or because this challenge was once run in 2015. We will remove those submissions. But we hope you enjoyed the experience :)
-
-
prashanb 5 years ago Hi I am trying out this problem with both ARIMA and SARIMAX but I am mostlys getting high predicted values where as expected out goes to <1000 after initial 8-9 days, anybpody else has faced this issue?
-
kennethleung_ty 3 years ago Likewise here. Did you manage to solve it?
-
-
smallet 9 years ago Thanks for this challenge. It was very educating. Is there a chance we will see editorials on the last three problems? I would be very pleased to learn the ways to solve them.
-
AffineStructure 9 years ago What I suggest you do is go to the Leaderboard of each problem. since the event is over, you can see how each person solved it.
you can also filter by programming language.-
smallet 9 years ago All those who got 50 points hardcoded the number. The next score is 35.38. It would be interesting to know how to solve this problem and get the right answer though.
-
-
PRASHANTB1984 9 years ago Do check the resources tab! Not a full solution, but it discusses various approaches.
-
-
mgbrooks208 7 years ago Hi everyone,
I am attempting this challenge but am having problems with the interface. I can't event begin to try to problem because I keep getting this error.
Traceback (most recent call last): File "/custom-RyLAs8OD9NUnDNn4D0Pu/solution.py", line 174, in <module> run_custom_checker(t_obj, r_obj) File "/custom-RyLAs8OD9NUnDNn4D0Pu/solution.py", line 76, in run_custom_checker outputs = map(float, outputs) ValueError: could not convert string to float:
This is without evening trying to load int the dail or peform any operation, only importing modules.
On a side note, I was also having an issue trying to import matplotlib.pyplot but was given an error that the module didn't exist?
Thank you for the help.
-
jjinking 9 years ago Anybody use regression models? I used previous x days as features, as well as the target's day of the week, but I'm not really getting any good results.
-
AffineStructure 9 years ago Regression models should not be a good fit here.
With time series we expect to see autocorrelation. Did you attempt to use GLM?
-
jjinking 9 years ago Haha yeah. Didn't do well.
-
-
Sort 11 Discussions, By:
Please Log In in order to post a comment