Guide: How to Use and Features (ver.1.03)

5. Making predictions TOC

• Submitting the data for prediction
• Accurracy of the predicted values
• Novelty degree

The "Prediction" page looks like the figure below.
You get some key information on the left, like ANN name, description and the names you chose for the input.
Additionally, some hints and a link that lets you open the analysis page in a separate window with the analysis for the current ANN.

The input questions are lines of data values, one for each variable, in the same format used in the beginning for the input data. Once again you can simply copy and paste values from a tab or space delimited text file or any spreadsheet. Apart from having as many input variables (columns) as the input data-set used to train the ANN there is no other restriction.
As you can see in the figure on the left, the values don't have to be aligned.

After pressing the "Predictions" button, you get the same page, but with the results appended.
Here is an example:

Input values (X) a b random1 random2 1 0 0 0 0 2 0 1 0 0 3 1 1 0 0 4 1 2 0 0 5 1 2 0.5 0.5 6 2 1 0.5 0.5 The input values you submited are listed first. Here you can check if you got it right.
Predicted values (Y=ANN(X)) a+b a-b 1 0.03861961825 0.153212212 2 0.4536942208 -0.8694173207 3 1.68677382 0.1347563292 4 2.975627916 -0.9059630096 5 3.108266518 -1.01921652 6 3.118425436 1.030767481 Here are the predictions listed, in the same order as the inputs and in the same order as you submited.
Confidence intervals of individual predictions (Y), for 95% probability level a+b a-b from to from to 1 -0.1338970951 0.190314182 0.03034468378 0.3279847404 2 0.2811775074 0.6053887845 -0.9922848489 -0.6946447922 3 1.514257107 1.838468384 0.011888801 0.3095288576 4 2.803111203 3.12732248 -1.028830538 -0.7311904811 5 2.935749805 3.259961082 -1.142084048 -0.8444439912 6 2.945908723 3.27012 0.9078999525 1.205540009 Here you can see how much you can trust the prediction.
Novelty index of individual input cases (X) submitted a+b a-b 1 out of range out of range 2 1.215493415 1.215493415 3 1.215493415 1.215493415 4 0.9507876864 0.9434864764 5 0.948175532 0.9468605442 6 0.9459535932 0.9469472215 Note: Values up to 1 indicate average novelty, the unlikelyhood of the data can be approximated for higher values by the one-tailed normal distribution applied to z=(novelty index-1). For example, 95% unlikelyhood is approximately 2.6 novelty. MD5 checksum of trained network: 0b50d5e62e47122884c0ebbd671908d2	Check the novelty of your question. Values clearly higher than 1 indicate values in your question very different from anything used to train the ANN. This is the case for the first query, where the novelty index is out of range.