Systems and methods for knowledge discovery in spatial data

• US 6,865,582 B2
• Filed: 01/02/2001
• Issued: 03/08/2005
• Est. Priority Date: 01/03/2000
• Status: Expired due to Fees

• Current Assignee: Battelle Energy Alliance LLC (Battelle Memorial Institute)
• Original Assignee: Bechtel BWXT Idaho, LLC
• Inventors: Obradovic, Zoran, Lazarevic, Aleksandar, Fiez, Timothy E., Vucetic, Slobodan, Pokrajac, Dragoljub, Hoskinson, Reed L.
• Primary Examiner(s): Vu, Kim
• Assistant Examiner(s): HAMILTON, MONPLAISIR G
• Application Number:
  US09/753,363
• Publication Number:
  US 20020038307A1
• Time in Patent Office: 1,526 Days
• Field of Search: 707/10, 707/104.1, 707/4, 707/2, 707/102, 702/2, 702179- 82, 702/5, 700/28, 706/904, 706/912
• US Class Current: 1/1
• CPC Class Codes: G06F 18/40 Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
• CPC Class Codes: Y02A 40/10 in agriculture
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing
• CPC Class Codes: Y10S 707/99948 Application of database or data structure, e.g. distributed, multimedia, or image

Abstract

Systems and methods are provided for knowledge discovery in spatial data as well as to systems and methods for optimizing recipes used in spatial environments such as may be found in precision agriculture. A spatial data analysis and modeling module is provided which allows users to interactively and flexibly analyze and mine spatial data. The spatial data analysis and modeling module applies spatial data mining algorithms through a number of steps. The data loading and generation module obtains or generates spatial data and allows for basic partitioning. The inspection module provides basic statistical analysis. The preprocessing module smoothes and cleans the data and allows for basic manipulation of the data. The partitioning module provides for more advanced data partitioning. The prediction module applies regression and classification algorithms on the spatial data. The integration module enhances prediction methods by combining and integrating models. The recommendation module provides the user with site-specific recommendations as to how to optimize a recipe for a spatial environment such as a fertilizer recipe for an agricultural field.

Claims

1. In a system including spatial data for a spatial environment, wherein a recipe is used in the spatial environment, a method for mining the spatial data to optimize the recipe for one or more target values, the method comprising:

an act of generating a data set from the spatial data using identified attributes selected by a user, the data set being varyingly complex based upon the identified attributes selected by the user;

an act of partitioning the spatial data into a training set and at least one modeling set wherein the act of partitioning is selected from the group consisting of;

selecting the training set such that the training set comprises a substantially homogenous spatial relationship to the at least one modeling set, and selecting the training set such that the training set comprises a substantially separate spatial relationship to the at least one modeling set;

an act of inspecting the generated data set to provide statistical information for the data set;

an act of preprocessing the data set to prepare the data set for modeling;

an act of modeling the preprocessed data set to describe relationships between the attributes and the one or more target values; and

an act of providing recommendations such that the recipe is optimized.

2. A method as defined in claim 1, wherein the act of preprocessing the data set further comprises:

an act of cleaning the generated data set;

an act of interpolating the generated data set;

an act of normalizing the generated data set; and

an act of generating new attributes.

3. A method as defined in claim 1, wherein the recipe is a fertilizer recipe for use in an agricultural field.

4. A method as defined in claim 1, wherein a crop yield is included in the one or more target values.

5. A method as defined in claim 1, further wherein the relationships include relationships between the training set and the at least one modeling set, wherein the training set is'"'"' used to optimize a recipe for the at least one modeling set.

6. A computer program product having computer executable instructions for executing the acts recited in claim 1.

7. In a system including one or more spatial databases corresponding to one or more spatial environments, a system for knowledge discovery from the one or more spatial databases, the system comprising:

a user interface; and

a spatial data modeling and analysis module (SDAM module) for extracting knowledge from the one or more spatial databases, the SDAM module comprising;

a data generation and manipulation module for loading a data set from the one or more spatial databases based on designated attributes, wherein the attributes are selected and supplied to the data generation and manipulation module by a user through the user interface;

a data inspection module for providing spatial statistics on the loaded data set;

a data preprocessing module for preparing the loaded data set for modeling, wherein the data preprocessing module removes errors from the loaded data set;

a data partitioning module for dividing the loaded data set into a training set and at least one modeling set wherein the dividing is selected from the group consisting of;

dividing such that the training set comprises a substantially homogeneous spatial relationship to the at least one modeling set, and dividing such that the training set comprises a substantially separate spatial relationship to the at least one modeling set; and

a modeling module for describing relationships between the attributes and one or more target values, wherein the relationships are obtained from the training set and applied to the at least one modeling set.

8. A system as defined in claim 7, wherein the SDAM module further comprises an integration module for enhancing the knowledge generated from the one or more spatial databases.

9. A system as defined in claim 7, wherein the preprocessing module further comprises:

a cleaning and filtering module for removing duplicate data and removing noise from the loaded data set;

a data interpolation module for computing common values for a common set of locations;

a data normalization module for transforming the loaded data set to a normal distribution and for scaling the loaded data set to a range;

a data discretization module for use in modeling the loaded data set;

a generating new attributes module for combining existing attributes into a single attribute;

a feature selection module for reducing the attributes identified by a user such that irrelevant attributes may be removed; and

a feature extraction module for reducing a dimensionality of the loaded data set.

10. A system as defined in claim 7, further comprising a recommendation module, wherein the recommendation module optimizes a recipe for a spatial environment.

11. A system as defined in claim 10, wherein the recommendation module includes at least one of:

a fertilization module for optimizing a fertilizer recipe to be applied to an agricultural field;

an irrigation module for optimizing a water recipe to be applied to a field; and

an equipment module for optimizing a recipe to be applied to equipment.

12. A system as defined in claim 11, wherein the recommendation module includes at least one of:

a pesticide module, a herbicide module, and a seed-spacing module.

13. A system as defined in claim 7, wherein each of the data generation and manipulation module, the data inspection, the data preprocessing module, the data partitioning module, and the modeling module can be independently controlled by the user through the user interface.