FAIL (the browser should render some flash content, not this).


                   MM-Data Mining | How It Works | MM-Data Mining Tools


MM-Data Mining

"Data visualization is especially useful for portraying fuzzy patterns that defy formulaic quantitative specification and for providing an overall qualitative feel for a dataset... Graphical displays should induce the viewer to think about the substance, present many numbers in a small space, make large datasets coherent, encourage the eye to compare different data, and reveal the data at several levels."

Edward Tufte, The Visual Display of Quantitative Information

MM-Data Mining technology delivers strategic topsight over enterprise data: understanding of the evolving big picture and instant drill-down ability to near real-time information. MM-Data Mining software conveys a higher bandwidth of information than 2-D and 3-D graphs or static charts, presenting information using color, shape, size, motion, and other dynamic visual elements. These decision support tools help executives look for problems and opportunities, track best and worst performing products, compare performance indicators in relation with other comparative measures and forecast trends to meet year-end goals. In this way, patterns of influence and variation are revealed, enhancing business situation awareness so decision makers can identify issues before problems occur and seize the initiative.

Integrating METAPHOR MIXER technology with user-specific Search Agents and best-of-breed OLAP/MDM engines delivers the capability to mine information along multiple dimensions, then fuse and visually display the results. MM-Data Mining systems help decision makers capitalize on information provided by OLAP and MDM techniques, such as how Profits are influenced by Location by Week, or how Sales vary Geographically across Product lines during a window of Time.

Information Landscapes form the infrastructure of MM-Data Mining decision support visualizations. Information Landscape schemas extend current MDM principals by allowing user-selected focus dimensions of the data being modeled to dictate the Landscape's structure. Landscapes assimilate business dimensions, attributes, and facts in a broad and immediately comprehensible visual grammar. They are navigable, interactive and inhabited by software agents programmed to perform various knowledge management and level of confidence functions.


MM-Data Mining knowledge tools help decision makers:

    • Slice information into multiple dimensions and present information at various levels of granularity.
    • View trends and develop historical tracers to show operations over time.
    • Produce pointers to synergies across multiple enterprise dimensions.
    • Provide exception analysis and identify isolated (needle in the haystack) opportunities.
    • Monitor adversarial capabilities and developments.
    • Create indicators of duplicative efforts.
    • Conduct What-if Analysis and Cross-Analysis of Variables.


How It Works

Multi-dimensional modeling (MDM) is a decision support architecture applied to data warehouses. This technique conceptualizes business models as a set of measures described by ordinary facets of business and views information from the perspective of a "slice of time". Unlike OLTP systems, which record discrete events or transactions such as journal entries, purchase orders or billing items, MDM systems are concerned with the quantitative results of events at intervals in time. Three concepts are fundamental to understanding MDM: Dimensions, Attributes, and Facts.

Dimensions are broad classes of business processes and functions. Common Dimensions include: Time, Geography, Company, Customer, Product, and View/Scenario. Dimensions --which can be broken down into smaller categories called Attributes -- provide the structure for the exploration of Facts

Attributes are focused classes or subsets of Dimensions. They provide the depth of Dimensions beyond identifying codes. For example, Attributes of a Product Dimension might include a hierarchy of Item, Class, Brand, and Division.

Facts are the numerical measures of the business process being modeled. These measures describe numeric computations aggregated by OLAP engines. Types of Facts include: Gross Sales, Percentage Gain/Loss, Transaction Volume, etc.

Utilizing MDM techniques, relationships and patterns are revealed by the existence of Facts at the cross section of Dimensions. For example, if there are sales from Customer X of Product Y, then a relationship between Customer and Product is implied.

Facts and Dimensions are physically represented in a relational database called a "star schema". Each Dimension in the star schema is described in its own table. Facts are arranged in a single large table indexed by a multipart key composed of the individual keys of each Dimension. Fact tables contain the quantitative data about a business and can consist of many columns and millions of rows. Dimension tables are smaller and hold descriptive data that reflect the Dimensions of the business. They describe Facts. For example, if a Fact is "Sales", Dimensions might be "Time, Geography, Customer, or Product".

MM-Data Mining supports the reporting and analytical needs of knowledge workers and executive decision makers by visually presenting numerical and categorical information in a Dimensions and Fact-based environment. Information Landscape schemas let decision makers interactively select Dimensions from the OLAP/MDM star schema and visually represent and explore them in relation to one another.

The categorical X and Y-axis of the Information Landscape schema are comprised of any two Focus Dimensions. These Dimensions provide the structure for the exploration of Facts. Facts -- the numeric measures of the business process being modeled-- comprise the Z-axis. For example, an icon's color and height above the Information Landscape indicates discrete, Product Sales information.

3-D icons represent the cross section of Dimensions and Facts. Icons of different shape, color, texture and behavior visualize the implicit relationship modeled by MDM at a given level of granularity. Hundreds or even thousands of discrete icons populate an Information Landscape in a structured, scalable manner.

Suppose an executive decision maker has the following query:

"Net Sales, in Dollars and Units, by Company for the last three years, as compared with Shipments and Budgeted Sales for the same period?"

This query provides us with a business process (Retail Sales Data), Facts (Sales Dollars, Sales Units, Return Units, Shipment Units, Price), and Dimensions (Customer, Product, Time, and View). Attributes (Sales, Shipments By Week and Store) are worked out by evaluating the more detailed features of the Dimensions at various levels of granularity or detail.

The MM-Data Mining interface for this query can be diagrammed as follows:

Product Lines occupy the (X) axis of the Information Landscape schema. Geographic Sales Regions are depicted along the (Y) axis. 3-D icons depict different Products (SKUs) within a Product Line. Height of the icons above or below the resulting grid (Z axis) represents Sales Volume or Volume Changes. Color and other visual characteristics (spinning, flashing) encode discrete numerical, as well as outlier statistics concerning attributes, making patterns for both extremes of opportunity and concern easily recognizable. Certain icons flash to indicate which attributes of the Products in what Sales Regions have been/are meeting Projected Sales Goals. Arrows associated with each icon indicate a change in the business' overall performance level, measured daily, weekly or monthly. These arrows clearly depict outlier dimensional attributes as they intersect Sales Facts.

Information Landscape schemas can be generated on the fly and can be manipulated to show, for example, criteria data, development trends, research gaps, duplicative efforts, and synergistic applications. Outputs can be used to document real and projected performance within specific areas to support and justify the funding activities.

Composite Information Landscape schemas (CLS's) aggregate landscapes hierarchically for scalability. CLS's provide users with enhanced situation awareness: the ability to quickly review the state of the enterprise and to keep track of progress being made based on variable criteria. More important, synergies of efforts can be detected and recommendations made about their integration. Since the entire enterprise can be presented within a CLS, decision makers can determine if strategic objectives are being met.


MM-Data Mining Tools

MM-Sales Analysis. A decision support system for managers in the retail/distribution industry. This product monitors all aspects of sales and indicates a business' opportunities and problems by tracking important dimensions such as products, subproducts, location and sales representatives. It provides an overview of a company's business by letting the user choose, measure and monitor a company's vital factors, such as: sales level; pricing; actual versus expected goals; global and detailed level indicators and geographies. Drill down to precise text and numbers on any given store or other granular level in the hierarchy with a mouse-click.

MM-Retail Banking. A decision support system for managers in charge of medium to large-sized, geographically dispersed retail banks. This product offers in-depth analysis of pre-selected retail indicators, navigation through information landscapes that are scalable to access multiple data sources. With a click of the mouse, indicators can be viewed geographically by areas, regions, territories, districts or specific locations.

MM-Fraud Detection. A decision support system for managers to track potential fraud activity. This product offers in depth-analysis of pre-selected fraud indicators, navigation through information and client/server access to scalable data warehouses. Comparisons and projections of key indicators reveal present and future trends in addition to spotting opportunities and problems. By maintaining these triggers, an executive will be proactive rather than reactive to fraudulent transactions.

MM-Credit Scoring. Evaluates credit-risk scoring parameters on hundreds of thousands of consumer loans. Credit performance can be depicted by location and other dimensions. In this way, the efficiency of different bank credit policies can be compared. Drill down helps the user identify the characteristics of similarly performing loans, as well as determine the most accurate measures for specific regions. A default view utility enables users to set preferences which render critical information and reveal relationships in credit performance patterns.

MM-Healthcare. Decision support tools for corporate benefits professionals that graphically render health industry data as three-dimensional information landscape, within which users navigate and interact. Programs compare the costs and benefits of HMO's, hospitals and doctors nationwide, breaking them into their component cost centers and clinical outcomes; or, provider practice - patterns by peer group and by national benchmarks.

MM-Bioinformatics. A decision support and investigative tool for combinatorial chemistry, high throughput screening, bioinformatics and genomic databases. In these areas, finding the needle in the haystack significantly enhances the product development cycle. Visual database mining tools are ideal for helping navigate through vast amounts of genomic data in the quest for new information about human disease.

MM-Image Event Detection (IED). A decision support system for analysts to track potential enemy activity within a 10 by 10 Km square digital image. This product offers in depth-analysis of pre-selected event indicators, navigation through image data and client/server access to scalable digital images. Comparisons of key indicators reveal present and future trends in addition to spotting events. By maintaining these triggers, defense forces will be proactive rather than reactive to IED or rocket attacks.