Wednesday, May 18, 2005

Solution: Decissive Predicting from historical data

The topic of this post really sounds weird.

As noted from my previous posts, Sampath has just managed to score 210.79 as cut-off. Currently we are not able to determine where will he be enrolled? To which college, to which stream. Based on this prediction we would like to evaluate the alternatives.

We are really stuck at this point as we could not make an educated guess. Reason, historical data is unavailable for the prediction. It will be nice if we could know the last year admission results. A simple table of data would do..

Attributes of table: {Year, Rank, Cut-off, College, Course}

Given this table we should be able to predict reasonably what are his chances. I know there are lot other parameters that varies each year and alter the admission results. But to begin with these details should be sufficient.

How great it will be if we could take these data and put a (decision tree based) data mining model above this and derieve the current year admission trends given the cut-off marks. I would love to work on this solution if I am given the data.

Why not let me start with some dummy data? Simultaneously try getting the real data and plug-in to the system. I am also thinking with year after year historical data, the system should have sufficient genetic algos to update itself with the trends and give more and more accurate results.

No comments: