Products

As a world leader in optimization and analytics, Elder Research Inc. offers software solutions to enable your business to better leverage existing data, automate and optimize business processes, and make better decisions. Many of these tools are available as stand-alone software applications or as Java or C libraries.

For more information or to request a demonstration, please contact sales@datamininglab.com or call (434) 973-7673.

Application	Versions	Description
Content Similarity Engine	Cross-platform API	Content Similarity Engine is a cross-platform API that uses statistical methods to categorize a large collection of text at adjustable levels of granularity (e.g. document, paragraph, phrase). It employs three approaches: exhaustive, “all-pairs,” and locality sensitive hashing (LSH) to identify non-obvious sites of interest, links between text sources, and original sources of information based on content profiles. The included ERI Exemplar algorithm automates scoring, ranking, and filtering, to help an analyst find documents with similar content by designating a document of interest (the exemplar) instead of the terms for a query. The exemplar method can make progress without requiring an analyst to label cases (e.g., interesting/non-interesting) for model training. Single-user license: $18,000 USD Upgrades and Maintenance: $3,600/yr
Entity Extraction Engine	Cross-platform API	Entity Extraction Engine is a cross-platform API based on the Stanford NER, a statistical approach that uses conditional random fields (CRF). The engine performs tokenization, part of speech identification, and noun phrase identification. Single-user license: $18,000 USD Upgrades and Maintenance: $3,600/yr
Collocation (Associative Network) Engine	Cross-platform API	Collocation Engine: ERI’s advanced collocation processor is a statistically-based predictive model contained in a cross-platform API able to identify the collocations within a document (e.g., identifying “United States of America” as a single entity or concept). The processor can be trained on example text from a single language, allowing it to “learn” statistical correlations within that language. The processor can then predict new collocations from this knowledge with great success and can do so in any (alphabet-based) language in which it is trained. Associative Network Engine: ERI’s engine automatically groups words into noun phrases (for example: “Washington, DC” and “assault rifle”) and generates a web of linked terms (similar to a thesaurus) from free text using a statistical approach (quite different than many other commercial tools that require pre-built dictionaries). It can learn, for example, that “AK-47”, “Kalishnikov”, and “Kalashnikov” are all the same concept – and highly related to the phrase “assault rifle”). Concepts are connected beyond immediate linguistic relationships, and extend to relationships based on higher-level meaning. Single-user license: $15,000 USD Upgrades and Maintenance: $3,000/yr
Multi-Dimensional Clustering Engine (Visualization)	Cross-platform GUI (2D and 3D) Cross-platform API	Multi-Dimensional Clustering Engine (Visualization): Insight is often enhanced by useful visual representations of complex data. Visualizations of high-dimensional, voluminous data allow an analyst to use the unmatched human strength of visual pattern identification. Unlike machines, analysts can identify an interesting new pattern or anomaly and quickly drive the focus to those patterns and clusters. From the highest visualization level, the analyst can rapidly drill down into key data while also identifying nearby clusters of similar sites or documents that might be worth exploring due to their proximity within the visualization. The 3D spring graph is interactive in practice, allowing an analyst to select the cluster of interest and drill into it by zooming or by exporting a selected group to a textual list of documents for further investigation. Single-user license: $15,000 USD Upgrades and Maintenance: $3,000/yr
Tagger Engine	Windows-only GUI	Tagger Engine: The ERI Tagger is a Windows GUI that allows for quick tagging, validation, and storage of the tagged results in an XML file. This technology can also be integrated with other engines to capture analyst feedback to allow models to adapt over time. Single-user license: $6,000 USD Upgrades and Maintenance: $1,200/yr
QuiltMaker	Cross-platform GUI	QuiltMaker is a visual software tool that quantifies association strengths hidden in transaction data. A key use is to provide “next most likely to buy” product information to sales representatives, allowing them to rapidly identify cross-selling opportunities. These association matrices, built from past transactions, are colored according to their value on user-selected attributes (hence the customer-coined name of “quilts”). QuiltMaker can also sort the rows and columns and re-color cells to identify strong product correlations. Products grouped into families can be imported so that individual products can be compared against other product families. Single-user license: $15,000 USD Upgrades and Maintenance: $3,000/yr
GROPE and Associated Optimization Algorithms	Cross-platform API	GROPE (Global R^d Optimization when Probes are Expensive) is a highly-efficient algorithm for global optimization that employs all known results to minimize the number of probes required. It is ideal for exploring functions with multiple modes, nonlinear or rough surfaces, a moderate number of dimensions, and an expensive evaluation function -- such as those arising from complex computer simulations. Unlike local search algorithms, it can find the overall global best, even if the surface has extreme changes and multiple local minima. GROPE also estimates the chance of improving one’s result with further probes, which is useful for planning experiments and knowing when to stop. This library also includes RandomSearch and GridSearch and a Conjugent Gradient Search (Powell’s Algorithm). The routines are in “advisor” rather than driver mode, to facilitate use with large and complex simulations.
Texas Two Step	Cross-platform GUI Cross-platform API	Texas Two Step is a proprietary decision tree algorithm that can select splits by looking one or two steps ahead, rather than merely one as with other tree algorithms. This usually leads to a more accurate decision model. The product induces models, selects variables, visualizes the tree, reports on training and evaluation errors and structures, handles categorical and real inputs and outputs, and can export the models created in PMML for further use by other commercial data mining tools.
Nearest Neighbor	Cross-platform GUI Cross-platform API	Nearest Neighbor is a graphical software tool that builds predictive models, using a consensus of k training cases that are most similar (are nearest neighbors) to the evaluation cases selected. The program employs Principle Components dimension reduction to display the results as an interactive scatterplot. A key feature is its automated discovery of the best combination of dimensions to employ in calculating nearness – the most important issue for such case-based algorithms. Estimation model results may be exported through PMML for use in other data mining tools.
ERI Entity Extraction Editor (E4)	Cross-platform GUI	E4 enables an analyst to correct the links and entity definitions induced by commercial entity extraction tools. Errors in extraction are unfortunately numerous – since “meaning” is often extremely difficult to automatically extract from text – but such errors can have deadly effect on downstream automated analysis. E4 can be used to clean up the links induced for high-value documents, making them much more useful for subsequent automated pattern discovery. The tool allows one to move, delete, and merge links that have been identified, and write new XML output file matching that produced by commercial tools (though now corrected). The modified files can be saved with a browse-able audit trail as XML or text, which can then be used to identify incorrect classifications. E4 also enables the user to visualize the model structure, graphically lay out the model, and export or print the graph image.