dataPOWER® dP-Mining Tools

dataPOWER mining tools are database-driven analysis tools designed to apply high-level analysis models that uncover hidden information in large data sets and in data sets spanning extended time periods.

Commonality (Included in dataPOWER core product)
Commonality Analysis is used to search for tools and other variables (date/time, recipe, operator, etc.) which may be responsible for good or bad results, such as poor yield. This analysis is based on a group of steps from lot history data. Commonality Analysis is also implemented in two versions: Database and Worksheet-based.

Database
• Fully integrated with dataPOWER, allowing data to be pulled directly from the database.
• Scheduled/batch processing.

Worksheet-based
• Applicable to data from any source.

{Click to see some examples}

Tool-to-Tool (Included in dataPOWER core product)
The Tool-to-Tool exception reporting utility identifies instances when the use of a particular tool (piece of equipment) at given steps in the production chain has a direct impact on parametric values, and therefore on yield. This is accomplished by taking a group of lots or a specified date/time range, and examining the tools that were used on that group, using lot history data and parametric tests such as wafer sort or PCM.

{Click to see some examples}

Classification Trees
The dataPOWER Classification Tree Engine (CTE) implements a state-of-the-art algorithm designed to identify common characteristics and likely causes of class membership in large-scale, long-term data sets. For example, when a large group of wafers can be divided into two categories — low- and high-yield — CTE can simultaneously consider the equipment used at each processing step as well as all metrology and electrical test results, looking for tools and/or parameters most likely to lead to yield success (or yield failure). The two implementations of Classification Trees are database and worksheet-based:

Database:
• Fully integrated with dataPOWER, allowing data to be pulled directly from the database.
• Scheduled/batch processing.
• Report storage/management.
• Multi-threaded.
• Scalable to very large databases.

Client-based:
• Applicable to data from any source.
• Reports storable as XML files.
• Supports interactive classification manipulation and generation.

Database & Worksheet-based Common Benefits
• Data drill-down from results for further analysis or verification.
• Robust management of missing and noisy data.
• Supports continuous and categorical attributes.
• Data sampling.

{Click to see some examples}

Self-Organizing Maps (SOM's)
A graphical representation method for forming overviews of multivariat data sets. Representation of a set of objects by a set of points in a low-dimensional space, so that objects that are similar to one another are represented by points that are close together (topology preserving mapping).