StarProbe Data Miner for Enterprise Applications

Enterprise data mining and intelligent analytic tools!

Corporate and government databases contain huge data (say, in tens or hundreds of millions of objects) that may consist of many variables, say, hundreds of variables including both categorical variables and numerical variables, with noisy information. Mining such data efficiently and effectively is always a serious challenge. StarProbe was specially designed to take on these challenging problems. It can help you to improve your performance on data mining and statistical analysis with exceptional leading-edge tools;

Rule-based Modeling and Rule-based Predictive Model Evaluation (NEW)
Rule-based Modeling Environment (RME) is a latest feature of StarProbe. It incorporates control rules and mathematical formulas with predictive modeling, providing most powerful modeling environment. RME is a unique platform which will definitely revolutionize your predictive modeling practice! For more, read the sections that follow!
Exclusively featured tools
Add these tools to your current data mining tool ranges! These tools can complement limitations of your current data mining tools!
Hotspot profiling & drill-down analysis, Cramer decision tree, Neural clustering, Cross table deviation analysis, Rule induction & class profiling, ...
Categorical dimensional modeling using star schema
Instead of simple single tables, star schema provides means to model input data in a way similar to corporate data warehouses and databases, combining multiple tables as single input data, which can lead to better data mining outcomes. For more, read star schema.
High performance on big enterprise data
Mining large enterprise data requires specially developed system for that end. StarProbe star schema supports complex categorical information very efficiently and reduces data file sizes, which in turn leads to greater performance. See our benchmark results.
64-bit precision
Both integers and reals are processed using 64-bit representation, allowing extremely large or small numbers can be computed with extreme-high accuracy.
Intuitive interfaces with color-enhanced graphical visualization
This makes StarProbe very easy to understand and to use for otherwise very complex advanced tools!
Smooth database/data warehouse integration
StarProbe works with databases seamlessly. It allows users to model and prepare star schema data interactively with databases. In addition, models can be developed and applied directly to database tables on the fly or later.
Multiple platforms and Multi-mode installation
The same copy of StarProbe runs on Windows, Mac OS X, Solaris, Linux, and so on. It can be deployed over the Internet or Intranet using HTTP web-servers and web-browsers. Alternatively, it can be installed as standalone application on desktops, giving greater flexibility in deployment.
Internet, Web Services, SOA ready!
Predictive models developed with StarProbe can be deployed over the Internet and Intranet easily using web services tools.

Data Mining - Full Range Algorithms

Data mining discovers hidden patterns in transactions and in objects (such as customers). StarProbe data miner supports full-range of data mining algorithms, including many exclusive pioneering features such as rule-based modeling, hotspot analysis, Cramer decision tree, and so on;

Rule-based predictive modeling and model evaluation (RME).
Gene (DNA) code pattern matching.
Neural network.
Decision tree and drill-down segmentation.
Hotspot analysis and profiling.
Rule induction.
Regression: General linear, logistic, and exponential.
Neural clustering / Self organizing maps (SOM).
Marketing segmentation.
Variable correlation link analysis.
Product association analysis.
Cross table & Deviation analysis.
Group-by summarization.
Visualization: Bar, pie, scatter plot, area, combo, ...
Statistics.

Hotspot Profiling & Conjoint Analysis

Hotspot analysis drills-down data systematically and detects important relationships, co-factors, interactions, dependencies and associations amongst many variables and values accurately, and generate profiles of most interesting segments. Hotspot analysis is very easy to use and provides analytic power un-paralleled by other statistical and OLAP tools. Hotspot Analysis performs the followings at the same time using Artificial Intelligence Technology;

Segmentation: Divide populations into segments.
Profiling: Develop profiles of hotspot segments.
Drill-down: Automatically drill-down dimensions and numerical value ranges.
Variable selection: Automatically select variables used in profiling and segmentation.
Ranking: Order segments based ranking criteria.
Visualization: Visualize result statistics.

Usage: profiling, customer profiling, risk factor profiling, segmentation, census analysis, survey data analysis, drill-down, database marketing, cross selling, and so on.

The following figure shows StarProbe hotspot profiling analysis module. From the module, you can drill-down automatically, analyze/compare details, perform gains/lift factor analysis, and segment design & analysis. For more, click Hotspot Profiling Analysis.

Decision Tree and Hierarchical Drill-down Segmentation

StarProbe data miner includes multi-purpose decision tree that can be used for classification, hierarchical drill-down, segmentation, statistical predictive scoring. It supports Cramer, Entropy, GINI, Chi-square, Twoing, etc.

Usage: risk modeling, direct marketing, database marketing, cross selling, and so on.

The following figures show StarProbe data miner decision tree drill-down and segmentation features. As you can see in the figures, users to identify interesting segments very easily using intuitive tree visualization. For more on drill-down and segmentation, read Decision Tree Software.

StarProbe data miner - drill down analysis.

StarProbe data miner - distribution analysis.

StarProbe data miner - segment & lift/gains analysis.

Predictive Neural Network

StarProbe data miner supports predictive neural network. StarProbe neural network is robust. Users can analyze predictive models through intuitive visualization features using easy-to-use graphical user interfaces. Models can be applied directly to database tables on the fly.

Usage: risk modeling, direct marketing, database marketing, and so on.

The following figures show StarProbe data miner neural network. As you can see in the figures, intuitive model visualization helps users to build and verify models easily, which makes significant difference in quality of models developed! For more, click Neural Network.

Neural Clustering and Statistical Predictive Segmentation Modeling

StarProbe Data Miner provides the most powerful clustering tool that combines with predictive modeling. StarProbe clustering is based on neural network known as Self Organizing Maps (SOM). SOM is how our brains learn and recognize patterns, features, and everything. Neural clustering works well on categorical information as well as numerical data. This has significant merits for business data. Note that business data are rich with categorical information. More importantly, neural clustering combines with RBF-style (Radial Basis Functions) predictive modeling, ideal for modeling statistically-oriented predictive problems. It allows you to model clustering as well as response behaviors, and apply to other data as predictive models.

Usage: segmentation modeling, customer segmentation, market segmentation, direct marketing, database marketing, cross selling, and so on.

The following figures show StarProbe Data Miner neural clustering. The top left figure shows a result of StarProbe neural clustering segmentation. Note that colors are used to denote segments. You can see how good segmentation neural clustering can produce! The top right is a gains chart for response & profit analysis for marketing. The figures of the middle row show variable-value distribution of segments. The left figure shows pie charts for a categorical variable. It shows all segments are group with single values. In addition, the same values are distributed nearby segments. The middle shows the similar patterns for a numerical variable. The right is all-in-one charts for a specific segment. The bottom figures shows variable-sensitivity to segments and gains/lift factor analysis. For more, click Neural Clustering.

Data Mining Tools: Visualization of segmentation.

Direct marketing response & profit analysis gains chart.

Data Mining Tools: Categorical data distribution.

Data Mining Tools: Numerical data distribution.

Data Mining Tools: Distribution visualization of cells.

Data Mining Tools: Segment sensitivity to input variables.

Data Mining Tools: Lift factor analysis and segmentation.

Marketing Segmentation and Modeling

This is a special tool for segmentation marketing. It combines with response and profit analysis. More importantly, this has modeling capability. Users can build response models based on past marketing results using information at the time of previous marketing campaigns and apply the models to the current information. An ideal tool for RFM and direct mail marketing!

Usage: segmentation modeling, RFM marketing, direct mail marketing, and so on.

Profile segmentation.

For more on segmentation marketing, click Direct Mail Marketing / Catalog Sales and RFM Marketing.

Visual Link Analysis - Correlation Analysis

StarProbe provides versatile visual link analysis tools. Analysis methods include correlations, associations by counts, averages, and totals. Correlation analysis detects various correlations using discretization, linearization, averaging and ranking automatically. In addition, powerful visualization allows you to analyze results very easily, with the help of color-enhanced wonderful graphics!

Usage: variable selection for segmentation and predictive modeling, medical & biological research, transport network analysis, and many other dependency analysis.

The following figure shows StarProbe link analysis tool. It shows an example of correlation analysis. The left panel shows correlation between variables. The right panel shows details of selected variables. Using various controls in the window, users can identify important relationships visually. For more or to see the figure of original size, click Link Analysis page.

Items/Products Link Analysis

Items link analysis can be used to analyze customers' product purchasing patterns. It can be used to identify not only co-products that customers buy together but also anti-products that customers tend not buying the other products.

Usage: Customer products purchasing patterns, items (objects) association analysis, etc.

The following figure shows StarProbe items link analysis tool. It shows an example of items link analysis. For more, click Link Analysis page.

Cross Tables and Deviation Analysis

StarProbe data miner produces versatile cross tables that incorporate totals, counts, arithmetic means, chi-square statistical analysis, percentages, hotspot analysis and 3D visualization all together at the same time. More importantly, deviation analysis, based on chi-square statistic, can uncover patterns hidden underneath cross table numbers.

Usage: dimensional data analysis, business performance analysis, deviation analysis, and so on.

Interpreting numbers of cross tables can be very confusing and misleading. StarProbe cross tables provide answers for this problem. The followings show relative importance of variations in numbers. You can easily identify business segments that under-perform or over-perform! For more, click Cross Table Analysis.

StarProbe Data Miner

Data Mining Methods

Data Mining Tools