Using Neural Networks and OLAP Tools to Make Business Decisions
A business analyst often wants to get a big picture of the business, to see broader trends based on aggregated data, and to see these trends broken down by any number of variables. Business intelligence (another term given many different definitions that I will not attempt to reconcile) is, in many instances, the process of extracting data from an OLAP database and then analyzing that data for information that you can use to make informed business decisions and take action.
Because the OLAP storage unit is multidimensional, it is called a cube as opposed to a table. With OLAP, the user directs the analysis and explores hypotheses or relationships. It is important to understand that OLAP requires the user to know what information he or she is searching for; that is, OLAP techniques do not process enterprise data for hidden or unknown intelligence.
Figure 3 illustrates the OLAP tool that's built into Microsoft Excel being applied to a data set that will be used throughout this article.
Figure 3: Excel 2007 with native OLAP tool (pivot table). OLAP cubes are designed for ad hoc reporting.
Although OLAP is able to provide numerical and statistical analysis of data in an efficient and timely way, it lacks the ability to make predictions.
In contrast, neural networks involve the automated process of finding relationships and patterns in data. For example, a company might want to know what pattern of behaviors predicts that a customer might leave for a competitor. Using computationally complex algorithms, the software finds relationships that were previously unknown.
Neural networks are capable of learning complex relationships in data. They can discern patterns in data, and then extrapolate predictions when given new data. The problems neural networks are used for can be divided in two general groups:
- Classification Problems: Problems in which you are trying to determine what type of category an unknown item falls into. Examples include medical diagnoses and prediction of credit repayment ability.
- Numeric Problems: Situations where you need to predict a specific numeric outcome. Examples include stock price forecasting and predicting the level of sales during a future time period.
When using NeuralTools, neural networks are developed and used in four steps:
- Data Preparation: The data you use in NeuralTools is defined in data sets. A Data Set Manager is used to set up data sets so they can be used over and over again with your neural networks.
- Training: With training, a neural network is generated from a data set comprised of cases with known output values. This data often consists of historical cases for which you know the values of output/dependent variable.
- Testing: With testing, a trained neural network is tested to see how well it does at predicting known output values. The data used for testing is usually a subset of your historical data. This subset was not used in training the network. After testing, the performance of the network is measured by statistics such as the % of the known answers it correctly predicted.
- Prediction: A trained neural network is used to predict unknown output values. Once trained and tested, the network can be used as needed to predict outputs for new case data.
Training and testing are an iterative, sometimes time-intensive process. Typically, you may train several different times with different settings to generate a neural network that tests best. Once you have your "best net," you can use it quickly for predicting.
Two Neural Network Examples with External Optimization: Auto Loans
NeuralTools can be used to predict unknown values of a category-dependent variable from known values of numeric and category independent variables. In this example, shown in Figure 4, a neural network learns to predict whether an auto loan applicant will be making timely payments, late payments, or default on the loan.
Figure 4: Excel 2007 with the NeuralTools and Evolver plugins applied to the same data set used in Figure 3.
The data set contains information on applicants who took car loans in the past, except for the first five rows, which contains data on new applicants. A neural net has been trained, with options to Automatically Test and Automatically Predict after training, as configured in the Application Settings dialog shown in Figure 5.
Figure 5: Configuration of Detailed Report, Type of Net, and so forth.
Page 2 of 6