iSixSigma

Prediction for future based on past data

Six Sigma – iSixSigma Forums Old Forums General Prediction for future based on past data

Viewing 11 posts - 1 through 11 (of 11 total)
  • Author
    Posts
  • #31140

    Seetharaman
    Member

    We would like to predict the pattern of any data based on the past data for 3 months of a particular parameter in our business.  Can any one help us to do this statistically?
    Thanks

    0
    #81890

    Michael Schlueter
    Participant

    Dear Seetharaman,
    I do think so; at least it something worth trying. What is quite easy to do is creating an early-warning system for a set of key-patterns: when your business pattern changes, you will know in advance. This system can be optimized for reliability.
    Anticipation of earth-quakes, epidemies, collision, work-load, business-situation etc. were practical applications of the technique I have in mind.
    I invite you to contact me via [email protected], when you are interested, so we can clarify more details.
    Regards, Michael Schlueter

    0
    #81904

    Chip Hewette
    Participant

    There is an old Andy Griffith episode where a new deputy takes all the past data and predicts that a certain law would be broken near midnight on a certain date.  The episode pokes fun at science and statistics, with good reason.  It is very difficult to make predictions!
    I assume your past data contains information that some would consider ‘predictors’ and ‘responses.’
    If you study the responses first, using a time-based view, you can evaluate if the responses have any sort of distribution.  An individuals and moving range chart is a good starting point.  Chances are the responses will be quite random, and have a wider distribution than first imagined.  If necessary, transform the responses to ‘difference from expected’ values to see if the differences are normally distributed. Understand your response data first.
    If you have many predictors associated with each response, some modeling could be useful.  However, modeling without good business process knowledge leads to erroneous conclusions.  Some predictors can actually be responses to business conditions, yet when put in the analysis as a predictor apppear very, very important.  (Lack of knowledge about the business leads some analysts to put data in the wrong category.)  Of course they would appear to be very important, as they are essentially the same as the response data.  Be very careful!

    0
    #82014

    Reinaldo Ramirez
    Participant

    There are some methodologies to forecast in the book Forecasting Methods for Management, Steven C. Wheelwright and Spyros Makridakis, John Wiley and Sons.
    ISBN 0-471-05630-8

    0
    #82017

    JoelB
    Participant

    Depends on the type of data you are analyzing. 
    You might want to look at books on Time Series analysis or Regression Analysis.

    0
    #82025

    aush
    Participant

    Try moving averages on Excel.

    0
    #82027

    njm
    Participant

    We use predictions often since our process is fairly long and we can’t wait to take action.
    It utilizes time, absolute values and past performance with regression in order to determine future results.
    It’s not an exact science — but better than finding out 3 months down the road that there is a major problem.

    0
    #82047

    Eoin
    Participant

    Send on your data and I’ll give it a crack. [email protected] please include as much commentary as you can on the context of the data.
     

    0
    #82053

    Magritzer
    Participant

    As I see ‘forecasting’, the technique or alorithm depends on the situation.  E.g., given the following series: 70, 72, 74, 76, 78, what is the next number?  If one uses a certain class of tools like moving average, exponential smoothing, simple linear regression, etc., then the intuitive (and “correct”) answer would be 80-ish.  However, ‘spoze that the series represented the monthly average temperature somwhere in Spain for April, May, June, July, August… it seems reasonable that the next month/number may indeed be less than 78 (September being cooler).  For that situation (and for situations like monthly tons of cement shipped, etc.) one would use a different technique (like classical/seasonal decomposition).  In short, one type of forecasting looks at [only] the data’s pattern itself; the other type incorporates multiple variables (as multiple linear regression would).  Within these two ‘categores’ there are subsets of techniques: sometimes moving average may be useful, sometimes it makes a rotten forecaster and adaptive filtering is better.    Like a lot of questions, the answer is, ‘it depends..’

    0
    #82058

    USM
    Member

    Wonderful consideration : “IT DEPENDS” ! ! !   In other words, for the seasonal example you¨re illustrating it would be perhaps wrong and/or absurd to forecast  a fourth month (September) with just three previous months history (June, July and August) as is required by Seetharaman at the start of this thread

    0
    #82078

    Michael Schlueter
    Participant

    Hi USM,
    Let’s put it that way: you need to know basic properties of your system. And you need to evaluate appropriate information.
    Example: I drive my car; its velocity varies like the zinc-price would do. Analysis, monitoring goes fine – until I crash. In retrospective I find a steep decrease of my velocity from high down to 0 within less than 10 s. Had I anticipated this dramatic change earlier – it would have been too late anyway.
    It would have been more appropriate had I measured the free distance ahead of me (well, this is how most of us do drive). Then early warning and appropriate action is both possible and reliable (within certain limits; think of mega-crashes).
    The question now is: what kind of system is the zinc-price-system? What are its basic properties, its fundamental laws we need to capture? What is the nature of the zinc-price? Cause or effect? X or Y? What do we need to know? In the driving analogy: is it more like the velocity? or more like the free distance measurement? or more like the variation in tire pressure during a high speed drive?
    If it is not the analog quantity to the free-distance-measurement we have to identify a different quantitiy in the zinc-price-system – and evaluate the relationship with zinc-price. Then winter can come ;-)
    Michael Schlueter

    0
Viewing 11 posts - 1 through 11 (of 11 total)

The forum ‘General’ is closed to new topics and replies.