Objective
In this challenge, we practice predicting time series. Check out the Resources tab for an overview of potential approaches to this problem.
Task
You are given the web traffic data for a particular website, which is measured in terms of user sessions. You are provided with the number of sessions for a time series of consecutive days starting from . Your task is to predict the number of sessions for the next days.
Input Format
The first row contains integer, .
Each of the subsequent lines contains an integer denoting the number of user sessions for day (where ).
Hidden Input File
The input file has rows (), each containing an integer.
The first integer is the number of sessions on .
The second integer is the number of sessions on . This pattern continues until we get to the last integer, which denotes the number of sessions on .
Sample Input File
The Sample Input file has rows () from the same data for consecutive days starting from . Your output for this data will be calibrated against the next days of data in the series, as predicted by you.
After inspecting the data (you may plot the first rows from the Sample Input), you may observe large periods of somewhat periodic and stable trends, followed by certain abrupt dips and jumps. The abrupt dips and jumps typically occur when there is a major change or revamp to the site content. You may assume that no such drastic change was made to the website after . So, any major trend observed for the first days of data will not be drastically disrupted in the next days for which you must predict the values.
Constraints
Scoring
The final score obtained upon submitting your code will be based ONLY on the hidden test case.
We will compute the mean of the magnitude of the percentage difference by comparing your expected answers with the actual sessions for each of the missing records in all test cases, samples included:
Your final score on a scale of will be:
If the mean value of exceeds (i.e. your predictions are off by or more on average), you will score zero. If your predictions are right on target, you will score .
When you hit Run Code (instead of Submit Code), we will run your solution against the sample test only. At that time, the visible score will be normalized out of rather than . In case your program throws an error (or has an incorrect output format) for a single test case, the overall score assigned will be zero.
Output Format
For each of the days of predictions, print your number of predicted sessions for that day on a new line. For example, the first line should contain the predicted number of sessions on , the second line should be the predicted number of sessions on , and so on.
Sample Input
1339
1462
1702
1656
1439
1208
1613
1935
1964
2003
2023
1559
1274
1805
2051
2024
2049
...
...
...
Sample Output
1000
1500
....
....
....
(30 such integers, each on a new row)