Let's say you have a table with historical stock price data, two technical analysis functions and you would like to get a prediction for tomorrow's Closing price:

Step 1. You click on PredictorXL from the menu in MS Excel

After the launch of the program you will see the form of PredictorXL.

Step 2. Using the mouse button choose the range of numerical data (Inputs) of the table that you want to use and which is influencing on the outputs and choose the range for the Outputs. We set the range of inputs to be all the data from B13 to H84. The outputs are specified as the values from E14 to E85.

Step 3. Align the range if necessary. Check "Scale input and output values" box if you would like the values of the range specified to be scaled. Neural networks require all input and output data to be in the allowable range of the activation function. This option scales the data to this range and when outputting the result the data is being scaled back. You can also set the number of epochs, minimum weight delta and choose whether to view the learning process.

Epochs. Epoch is a full cycle of neural network training on the entire training set. This parameter defines the maximum learning epochs cycles to reach specified minimum weights delta.

Minimum weight delta. Within the training process synapses weights are being corrected.  Minimum weight delta defines the desired lowermost weight correction value when the network is considered to be trained enough.

Step 4. Set the following parameters (for advanced users):

1. Initial weights: Initial weights of synapses. Synapses of every neuron will be initialized with random values from 0 to the initial weights.
2. Learning rate: a value between 0 and 1 that affects the rate at which the network learns. The larger the learning rate, the faster the network will converge. Be advised that oscillation and non-convergence may occur if the learning rate is set too high.
3. Momentum: High learning rates often lead to weight change oscillations during the training process, which may cause non-convergence or return of a non-optimal solution. Momentum makes it less likely for such undesirable cases to occur by making the next weight change a function of the previous weight change to provide a smoothing effect. The value for momentum (between 0 and 1) determines the proportion of the last weight change that is added to the next weight change.
4. Activation function: There are five functions available: Threshold, Hyperbolic tangent, Zero-based log-sigmoid, Log-sigmoid and Bipolar sigmoid.
5. Neurons in hidden layer: Set the number of neurons in hidden layer

Note: In most cases, the default values are acceptable for these parameters.

Step 5. Click the New button, followed by the Start training button. If Show learning process option is checked, a graphical representation of the training process is displayed. The green graph shows the actual values, and the red the predicted ones:

Step 6. When the training is complete, you should specify the input and output ranges for the prediction and press Predict button. The result is in E87 cell: