Selecting Input and Target Columns

By default, after data analysis the last column will be considered as the target (output) one and other columns will be considered as input columns. - target column is marked with light-pink background in the column header;

• Input columns are marked with light-blue header background.

To select the target column manually, click on the column header to select the column and then click on the Mark as Target button of the Data Analysis Toolbar or select the target column name in the Target drop-down box. The column you selected will be marked as target and the previous target column will be marked as input. In the same manner you can:

Accept / Reject

Change Column Type Note: You can select only one column as a target.

Understanding Data Partition

NeuroIntelligence divides each dataset onto three sets:

Training set- is a part of your input dataset used for neural network training, i.e. for adjustment of network weights.

Validation set is a part of your data used to tune network topology or network parameters other than weights.

Test set is a part of your input data set used ONLY to test how well the neural network will perform on new data. The Note: you may also decide to ignore some parts of your data if your dataset is too large or you want to experiment with smaller amounts of data. In this case one more set will be introduced - Ignored set. You dataset can be partitioned manually or automatically.

Partitioning Dataset Automatically

Click Partition button on the main toolbar to perform automatic data partition. You can specify parameters of the automatic partition in Partition Options.

Data preprocessing means modification of the data before it is fed to a neural network. Preprocessing transforms the data to make it suitable for neural network. Neural networks work only with numeric data. Often, dates and time as well as categories (textual data represented by a final set of values) need to be processed. Therefore, additional transformation of such data is required before feeding them to the network input (preprocessing) and after their reception at the output. Note: Textual values can be used only if they represent a final set of values. Columns with such values are called categorical columns

Working with Preprocessing Window

To preprocess the dataset, click Preprocess on the main toolbar or select Data > Preprocess. You can also select the Preprocessing window using Stages tab at the bottom of the main window. Note: The dataset can be preprocessed automatically after data analysis (see Preferences for more details) In the Data Preprocessing Window you will see:

Encoded Columns - list contains column names for preprocessed columns. Column Details widow contains details for the selected column.

Encoded Data data-grid is filled with preprocessed data.

You can change parameters of data preprocessing in Data Preprocessing Options.

To design a network, you need to specify the network architecture ( number of hidden layers and units in each layer) and network properties (error and activation functions). Note: You can run one of built-in search methods. The search method tries to find the best design options automatically. In both cases NeuroIntelligence creates a feed-forward fully-connected neural network (multi-layer perceptron).