# Prediction based on previous historical data

22 views (last 30 days)
Mekala balaji on 6 Feb 2015
Answered: KUMANAN KUPUSAMY on 28 Jan 2017
Hi, I have Historical data of "Start" and "End". Now, I have "Start" value, but want to predict "End" value. Can some one help me how to predict today's End value based on available Historical "Start & End" and today's Start value (to Predict Today End value). The day is of random type no correlation with time. Please help me some available models using MATLAB. Many many thanks in advance.
Historical Data:
Start End
0.250 0.256
0.240 0.240
0.224 0.240
0.260 0.270
0.240 0.260
And Today Start value: 0.21 (Today End Value: 0.23)
Sincerely, Your's

Guillaume on 6 Feb 2015
Edited: Guillaume on 6 Feb 2015
You don't have enough points to infer any kind of regression. I mean, if you look at:
data = [0.25 0.256
0.24 0.24
0.224 0.24
0.26 0.27
0.24 0.26];
scatter(data(:, 1), data(:, 2));
there's no pattern to your data.
If you want to do a linear regression anyway:
polystartend = polyfit(data(:,1), data(:, 2), 1); %linear regression
todaystart = 21;
todayend = polyval(polystartend, todaystart)
gives
todayend = 0.22688

#### 1 Comment

Mekala balaji on 7 Feb 2015
Sir Thank you so much.
I have upper & lower bound: 0.005 from previous end value. If I predicted today's end is greater than 0005, then I want to restore the just previous end value as today's end value. In this case 0.26 (just for example. but in this no problem). I just want to set this boundary condition (constraint).So, how can I set this condition.
Many many thanks.

Udit Gupta on 6 Feb 2015
You can use mtlabs Curve Fitting tool (cftool). If you don't have that search for any linear regression or linear interpolation implementation on file exchange.

KUMANAN KUPUSAMY on 28 Jan 2017
can i use current data to predict previous data?