Sequential Pattern Mining. Predictive Analysis. Data Mining
Date: October 20, 2013 09:40AM
great forum if i may add.
I have a problem that's been haunting me like a ghost.
Foremost i am not a programmer neither do i have the basic skill-set that most of the esteemed wizards in this forum possess.
nonetheless, it doesn't stop me from accepting i am a novice wanting to learn how to merge my ideas and machine language together.
I have numerous number of sheets of tabulated historical XYZ events datasheets with columns and rows, where each column also has sub-columns.
If you stare hard at these sheets from sheet 1 to sheet 200, various parametric pattern begins to appear that can be used to predict the outcome of events slated in future XYZ events datasheets.
This is possible by using my own developed parametric patterns which includes words and number repetition patterns, colour coding patterns, Single Alphabet patterns, Number range patterns etc.
Also making it easier is the fact that i have access to future xyz events datasheets list...basically i am trying to increase my success rate at using the past historical xyz events datasheets to help make informed predictions from the future xyz events datasheets list.
In the past, I have succeeded in identifying successful predictions trends using the above mentioned parametric patterns, but when i try to put them to use, it fails,it sometimes goes accurate, but accuracy is on a 1% out of 100%.....1% successful in real life and 99% successful on my computer screen.
A quick example is this:
There are 4 events in a month.
Events 1 and events 2 and events 3 having been actualized, now belong in the past hence i refer to them as historical XYZ events datasheets.
So that means we still have events 4 left in the month.
*Remember that i always have access to each subsequent future events xyz datasheets e.g
In our example above, Using my own developed parametric patterns, i can use the data from events 1 historical datasheets to discover sequential patterns in events 1 datasheets that also reappears in events 2 datasheets, furthering this trend, if i apply this same sequential patterns, i will discover same sequential patterns and results in events 3 datasheets although sometimes the results and sequences are distorted a little especially when it is getting to the 3rd events.
Anyway, furthering on some more, on the future xyz events datasheet for events 4, i will begin to discover same sequential patterns for events 1, events 2 and events 3 appearing on events 4 datasheets .....
From this discovery i aim to make informed predictions....now events 4's prediction on paper or on my computer screen always pans out successfully 60-100% but in reality( when i put my money on it) on a larger scale it always fails.
I am looking to increase my chances of discovering this sequential patterns on time, to make predictive decisions/results for events 2 and 3 as soon as i set my eyes on events 1...remember that i have access to subsequent future xyz events datasheets.....
So as soon as events 1 is actualized, i want to be able to use events 1's historical datasheets to filter out higher success rates of prediction from events 2 datasheets .....from experience i have realized that trying to be safe by waiting for repeat success of predicted results in events 2 and 3 before actually investing when events 4 arrives, almost always fails but on my computer screen it never fails, i dont want to wait till events 3 or events 4 again....i want to hit the mark on events 2.
My request is this.
Can rapidminer analyze past, present and future trends using the historical XYZ sheets i have and use this analysis to anticipate emerging trends by using existing patterns in the historical sheets i have?
Will it help if i provide parametric and sequential trends that rapidminer can use along with its own method of revealing patterns and signals from structured and unstructured information on my historical XYZ sheets?
(This parametric and sequential trends exists as patterns, repetition patterns, colour coding patterns, Alphabet patterns, Number range patterns etc.)
Can rapidminer intelligently filter and create results from a wide range of historical datasheets based on the parametric values that it will identify?
The parameters include the repetition patterns, colour coding patterns, Alphabet patterns, Number range patterns as well as information sought from the future XYZ events datasheets.
Basically, though the predictive data i want to discover exists in the future xyz events datasheets, i expect rapidminer (using or not using my suggested parameters) to process data from the historical xyz events datasheets to discover this predictive data from the future xyz datasheets and display the related data for me to make informed predictive decisions.
I hope this makes sense?
I have the events database in question and can provide it for your perusal if need be.
I only want to know if this idea of mine is possible to achieve.