# US 2006/0136462 A1 · Data-centric automatic data mining

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2006/0136462 A1
- Document ID: US20060136462A1
- Status: Published application
- Kind: A1
- Application number: 11/012,350
- Filed: 2004-12-16
- Published: 2006-06-22
- Earliest claimed: 2004-12-16
- Assignees: Oracle International Corporation
- Inventors: Marcos Campos; Boriana Milenova; Peter Stengard
- Web page: https://patentgrep.com/patent/US20060136462A1

## CPC

- G06F 16/2465: {Query processing support for facilitating data mining operations in structured databases}
- G06F 16/285: {Clustering or classification}
- Y10S 707/99931: Database or file accessing
- Y10S 707/99933: Query processing, i.e. searching
- Y10S 707/99935: Query augmenting and refining, e.g. inexact access
- Y10S 707/99956: File allocation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- Y10S: TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS

## Abstract

A data-centric data mining technique provides greater ease of use and flexibility, yet provides high quality data mining results by providing general methodologies for automatic data mining. A methodology for each major type of mining function is provided, including: supervised modeling (classification and regression), feature selection, and ranking, clustering, outlier detection, projection of the data to lower dimensionality, association discovery, and data source comparison. A method for data-centric data mining comprises invoking a data mining feature to perform data mining on a data source, performing data mining on data from the data source using the data mining feature, wherein the data mining feature uses data mining processes and objects internal to the data mining feature and does not use data mining processes and objects external to the data mining feature, outputting data mining results from the data mining feature, and removing all data mining processes and objects internal to the data mining feature that were used to process the data from the data source.

## Claims

### Claim 1

A method for data-centric data mining comprising:
invoking a data mining feature to perform data mining on a data source;
performing data mining on data from the data source using the data mining feature, wherein the data mining feature uses data mining processes and objects internal to the data mining feature and does not use data mining processes and objects external to the data mining feature;
outputting data mining results from the data mining feature; and
removing all data mining processes and objects internal to the data mining feature that were used to process the data from the data source.

### Claim 2 (depends on 1)

The method of claim 1, wherein the user does not need to specify data preparation to be performed on the data source, regardless of the data source.

### Claim 3 (depends on 2)

The method of claim 2, wherein the data mining features perform tasks based on needs and perspectives of the user without exposing to the user data mining specific terminologies and methodologies.

### Claim 4 (depends on 1)

The method of claim 1, wherein the data mining features comprise at least one of:
an EXPLAIN data mining feature that identifies attributes that are important or useful for explaining variation on a target attribute of interest and that provides a ranking of the attributes and a figure of merit;
a PREDICT data mining feature that produces predictions for unknown targets for classification mining tasks and regression mining tasks, and a figure of merit indicating the overall quality of the predictions;
a GROUP data mining feature that clusters similar records together and provides description for the groups;
a DETECT data mining feature that identifies records in the data from the data source that are atypical, outliers, or anomalies;
a PROJECT data mining feature that projects the data from the data source to a space of lower dimensionality and provides description of newly created attributes;
a DISCOVER data mining feature that uncovers hidden patterns in the data from the data source and provides a list of discovered patterns; and
a COMPARE data mining feature that compares content of two data sources and produces a summary of differences and similarities discovered.

### Claim 5 (depends on 1)

The method of claim 1, wherein the data mining features comprise at least one of:
a data mining feature that identifies attributes that are important or useful for explaining variation on a target attribute of interest and that provides a ranking of the attributes and a figure of merit;
a data mining feature that produces predictions for unknown targets for classification mining tasks and regression mining tasks, and a figure of merit indicating the overall quality of the predictions;
a data mining feature that clusters similar records together and provides description for the groups;
a data mining feature that identifies records in the data from the data source that are atypical, outliers, or anomalies;
a data mining feature that projects the data from the data source to a space of lower dimensionality and provides description of newly created attributes;
a data mining feature that uncovers hidden patterns in the data from the data source and provides a list of discovered patterns; and
a data mining feature that compares content of two data sources and produces a summary of differences and similarities discovered.

### Claim 6 (depends on 1)

The method of claim 1, wherein the data mining is performed by:
receiving data upon which to perform data mining processing;
computing statistics on the received data;
building a data mining model using the received data and based on the computed statistics;
determining a quality of results of the built data mining model;
selecting a data mining model based on the determined quality of results; and
applying the selected data mining model to the received data to generate applied data.

### Claim 7 (depends on 6)

The method of claim 6, wherein computing the statistics comprises:
computing a number of distinct values for each predictor attribute and the target attribute;
computing a number of total records included in the input training data;
computing bin boundaries when discretization is performed; and
computing shift and scale values for attributes where normalization is performed.

### Claim 8 (depends on 6)

The method of claim 6, wherein building a data mining model based on the computed statistics comprises:
determining an appropriate data mining function and algorithm to be used;
providing parameters to a build process of the data mining model so as to generate an optimal model;
selecting a data mining model based on the computed statistics; and
adjusting model creation parameters of the selected model based on the computed statistics.

### Claim 9 (depends on 6)

The method of claim 6, wherein determining a quality of results of the built data mining model comprises:
evaluating results of the built data mining models using a confusion matrix, for classification models; and
evaluating results of the built data mining models using computed measures of accuracy of the built data mining models, for regression models.

### Claim 10 (depends on 7)

The method of claim 7, wherein the computed measures of accuracy include at least one of a mean absolute error or a root mean square error, for regression models.

### Claim 11 (depends on 6)

The method of claim 6, further comprising:
preparing the received data before building the data mining model.

### Claim 12 (depends on 11)

The method of claim 11, wherein preparing the received data comprises at least one of discretization of continuous attributes, outlier handling, normalization of continuous attributes, missing values treatment for continuous and categorical attributes, and randomized sampling of input data to use for model building and testing.

### Claim 13 (depends on 6)

The method of claim 6, further comprising:
analyzing the applied data resulting from applying the selected data mining model to the received data.

### Claim 14 (depends on 13)

The method of claim 13, wherein analyzing the applied data comprises:
if the data mining model is a binary classification data mining model, performing receiver operating characteristics analysis of the applied data to generate predictions, and adjusting the predictions to support an optimal probability threshold resulting in a best trade-off between True Positive Rate and False Positive Rate.

### Claim 15 (depends on 14)

The method of claim 14, further comprising:
producing result data based on the analyzed applied data.

### Claim 16 (depends on 15)

The method of claim 15, wherein producing result data comprises:
generating a result table based on the analyzed applied data wherein an optimized probability threshold is used to determine a prediction.

### Claim 17 (depends on 6)

The method of claim 6, wherein the method is used to produce predictions for known and unknown targets.

### Claim 18 (depends on 17)

The method of claim 17, wherein the received data contains records where the target value is known and records where the target value is not known.

### Claim 19 (depends on 18)

The method of claim 18, wherein building the data mining model comprises:
building the data mining model using only the records where the target value is known.

### Claim 20 (depends on 19)

The method of claim 19, wherein applying the selected data mining model comprises:
scoring the data mining model using both the records where the target value is known and the records where the target value is not known.

### Claim 21 (depends on 1)

The method of claim 1, wherein the data mining is performed by:
clustering similar records together; and
providing description for the groups.

### Claim 22 (depends on 21)

The method of claim 21, wherein the clustering is performed using a pre-defined number of clusters or finding a number of clusters automatically.

### Claim 23 (depends on 21)

The method of claim 21, wherein the description includes at least one of:
a centroid described by a mean, for numeric attributes, or a mode, for categorical attributes;
a representative case for the cluster;
a histogram; and
a rule description of the cluster.

### Claim 24 (depends on 1)

The method of claim 1, wherein the data mining is performed by identifying records in the data from the data source that are atypical, outliers, or anomalies.

### Claim 25 (depends on 24)

The method of claim 24, wherein a measure of anomaly for each row is determined.

### Claim 26 (depends on 1)

The method of claim 1, wherein the data mining is performed by:
projecting the data from the data source to a space of lower dimensionality; and
providing description of newly created attributes.

### Claim 27 (depends on 26)

The method of claim 26, wherein a number of projections can be user specified or automatically determined.

### Claim 28 (depends on 1)

The method of claim 1, wherein the data mining is performed by:
uncovering hidden patterns in the data from the data source; and
providing a list of discovered patterns.

### Claim 29 (depends on 28)

The method of claim 28, wherein the list of discovered patterns is defined at an attribute level or at an attribute/value pair level.

### Claim 30 (depends on 1)

The method of claim 1, wherein the data mining is performed by:
comparing content of two data sources; and
producing a summary of differences and similarities discovered

### Claim 31 (depends on 30)

The method of claim 30, wherein the summary comprises discovered similarities, discovered dissimilarities, or both discovered similarities and discovered dissimilarities.

### Claim 32 (depends on 1)

The method of claim 1, wherein the data source comprises at least one of a relational table or view in a relational database, multidimensional (OLAP) data, a flat file (external tables), or a data streams.

### Claim 33

A method for data-centric data mining comprising:
attaching an existing data mining model to the data source;
invoking a data mining feature to perform data mining on a data source;
performing data mining on data from the data source using the data mining feature, wherein the data mining feature uses the attached data mining model;
outputting data mining results from the data mining feature; and
removing all data mining processes and objects internal to the data mining feature that were used to process the data from the data source, with the exception of the model attached to the data source.

### Claim 34 (depends on 33)

The method of claim 33, wherein performing data mining comprises building a data mining model on the fly at functional index construction time.

### Claim 35 (depends on 34)

The method of claim 34, wherein the data mining model is automatically updated as the data in the data source is changed.

### Claim 36 (depends on 33)

The method of claim 33, wherein performing data mining comprises using a data mining model that was built prior to invoking the data mining feature.

### Claim 37 (depends on 33)

The method of claim 33, wherein the user does not need to specify data preparation to be performed on the data source, regardless of the data source.

### Claim 38 (depends on 34)

The method of claim 34, wherein the data mining features perform tasks based on needs and perspectives of the user without exposing to the user data mining specific terminologies and methodologies.

### Claim 39 (depends on 33)

The method of claim 33, wherein the data mining features comprise at least one of:
an EXPLAIN data mining feature that identifies attributes that are important or useful for explaining variation on a target attribute of interest and that provides a ranking of the attributes and a figure of merit;
a PREDICT data mining feature that produces predictions for unknown targets for classification mining tasks and regression mining tasks;
a GROUP data mining feature that clusters similar records together and provides description for the groups;
a DETECT data mining feature that identifies records in the data from the data source that are atypical, outliers, or anomalies;
a PROJECT data mining feature that projects the data from the data source to a space of lower dimensionality and provides description of newly created attributes;
a DISCOVER data mining feature that uncovers hidden patterns in the data from the data source and provides a list of discovered patterns; and
a COMPARE data mining feature that compares content of two data sources and produces a summary of differences and similarities discovered.

### Claim 40 (depends on 33)

The method of claim 33, wherein the data mining features comprise at least one of:
a data mining feature that identifies attributes that are important or useful for explaining variation on a target attribute of interest and that provides a ranking of the attributes and a figure of merit;
a data mining feature that produces predictions for unknown targets for classification mining tasks and regression mining tasks, and a figure of merit indicating the overall quality of the predictions;
a data mining feature that clusters similar records together and provides description for the groups;
a data mining feature that identifies records in the data from the data source that are atypical, outliers, or anomalies;
a data mining feature that projects the data from the data source to a space of lower dimensionality and provides description of newly created attributes;
a data mining feature that uncovers hidden patterns in the data from the data source and provides a list of discovered patterns; and
a data mining feature that compares content of two data sources and produces a summary of differences and similarities discovered.

### Claim 41 (depends on 33)

The method of claim 33, wherein the data mining is performed by:
receiving data upon which to perform data mining processing;
computing statistics on the received data;
building a data mining model using the received data and based on the computed statistics;
determining a quality of results of the built data mining model;
selecting a data mining model based on the determined quality of results; and
applying the selected data mining model to the received data to generate applied data.

### Claim 42 (depends on 41)

The method of claim 41, wherein computing the statistics comprises:
computing a number of distinct values for each predictor attribute and the target attribute;
computing a number of total records included in the input training data;
computing bin boundaries when discretization is performed; and
computing shift and scale values for attributes where normalization is performed.

### Claim 43 (depends on 41)

The method of claim 41, wherein building a data mining model based on the computed statistics comprises:
determining an appropriate data mining function and algorithm to be used;
providing parameters to a build process of the data mining model so as to generate an optimal model;
selecting a data mining model based on the computed statistics; and
adjusting model creation parameters of the selected model based on the computed statistics.

### Claim 44 (depends on 41)

The method of claim 41, wherein determining a quality of results of the built data mining model comprises:
evaluating results of the built data mining models using a confusion matrix, for classification models; and
evaluating results of the built data mining models using computed measures of accuracy of the built data mining models, for regression models.

### Claim 45 (depends on 44)

The method of claim 44, wherein the computed measures of accuracy include at least one of a mean absolute error or a root mean square error, for regression models.

### Claim 46 (depends on 41)

The method of claim 41, further comprising:
preparing the received data before building the data mining model.

### Claim 47 (depends on 46)

The method of claim 46, wherein preparing the received data comprises at least one of discretization of continuous attributes, outlier handling, normalization of continuous attributes, missing values treatment for continuous and categorical attributes, and randomized sampling of input data to use for model building and testing.

### Claim 48 (depends on 41)

The method of claim 41, further comprising:
analyzing the applied data resulting from applying the selected data mining model to the received data.

### Claim 49 (depends on 48)

The method of claim 48, wherein analyzing the applied data comprises:
if the data mining model is a binary classification data mining model, performing receiver operating characteristics analysis of the applied data to generate predictions, and adjusting the predictions to support an optimal probability threshold resulting in a highest True Positive Rate versus False Positive Rate.

### Claim 50 (depends on 49)

The method of claim 49, further comprising:
producing result data based on the analyzed applied data.

### Claim 51 (depends on 50)

The method of claim 50, wherein producing result data comprises:
generating a result table based on the analyzed applied data wherein a probability threshold is used to determine an optimal prediction.

### Claim 52 (depends on 41)

The method of claim 41, wherein the method is used to produce predictions for known and unknown targets.

### Claim 53 (depends on 52)

The method of claim 52, wherein the received data contains records where the target value is known and records where the target value is not known.

### Claim 54 (depends on 53)

The method of claim 53, wherein building the data mining model comprises:
building the data mining model using only the records where the target value is known.

### Claim 55 (depends on 54)

The method of claim 54, wherein applying the selected data mining model comprises:
scoring the data mining model using both the records where the target value is known and the records where the target value is not known.

### Claim 56 (depends on 31)

The method of claim 31, wherein the data mining is performed by:
clustering similar records together; and
providing description for the groups.

### Claim 57 (depends on 56)

The method of claim 56, wherein the clustering is performed using a pre-defined number of clusters or finding a number of clusters automatically.

### Claim 58 (depends on 56)

The method of claim 56, wherein the description includes at least one of:
a centroid described by a mean, for numeric attributes, or a mode, for categorical attributes;
a representative case for the cluster;
a histogram; and
a rule description of the cluster.

### Claim 59 (depends on 33)

The method of claim 33, wherein the data mining is performed by identifying records in the data from the data source that are atypical, outliers, or anomalies.

### Claim 60 (depends on 59)

The method of claim 59, wherein a measure of anomaly for each row is determined.

### Claim 61 (depends on 33)

The method of claim 33, wherein the data mining is performed by:
projecting the data from the data source to a space of lower dimensionality; and
providing description of newly created attributes.

### Claim 62 (depends on 61)

The method of claim 61, wherein a number of projections can be user specified or automatically determined.

### Claim 63 (depends on 33)

The method of claim 33, wherein the data mining is performed by:
uncovering hidden patterns in the data from the data source; and
providing a list of discovered patterns.

### Claim 64 (depends on 63)

The method of claim 63, wherein the list of discovered patterns is defined at an attribute level or at an attribute/value pair level.

### Claim 65 (depends on 33)

The method of claim 33, wherein the data mining is performed by:
comparing content of two data sources; and
producing a summary of differences and similarities discovered

### Claim 66 (depends on 65)

The method of claim 65, wherein the summary comprises discovered similarities, discovered dissimilarities, or both discovered similarities and discovered dissimilarities.

### Claim 67 (depends on 33)

The method of claim 33, wherein the data source comprises at least one of a relational table or view in a relational database, multidimensional (OLAP) data, a flat file (external tables), or a data streams.
