Browse Prior Art Database

Fast Data Encoding For Spatial Association Rule Mining Online Service Disclosure Number: IPCOM000210696D
Publication Date: 2011-Sep-09
Document File: 5 page(s) / 148K

Publishing Venue

The Prior Art Database


This article presents a fast data encoding method for spatial association rule mining online service. The primary advantage of this new method over exiting ones is that it significantly reduces the time complexity and meanwhile guarantee the privacy. In addition, it also has the advantages of easier maintenance, easier deployment, and compatibility.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 23% of the total text.

Page 01 of 5

Fast Data Encoding For Spatial Association Rule Mining Online Service


Disclosed is a method enabling fast data encoding for spatial association rule mining online service. Online analytics service is becoming more and more important nowadays. Analytics-as-a-Service and public cloud computing are the future trend of analytics service. There are 2 issues to be considered in these scenarios: privacy and performance. The privacy issue is two-fold, including data privacy and result privacy. In a typical cloud computing scenario, user usually needs to upload data from client side to cloud side, and when analysis has finished and result has been generated, download the result back to client side. In such a scenario, the user's data is actually exposed to the cloud, as well as the generated result. The performance issue is mainly about service level agreement (SLA).

Spatial association rule (SAR) mining [1], which aims to find frequent spatial relationships between geospatial objects in spatial database, is an important analytics and can be offered as a powerful online service. SAR mining is geospatial extension of classical association rule mining [2-4] which aims to find frequent patterns like "beer=>diaper" in customer transactions. Instead of dealing with purchase transaction data, SAR mining is used to correlate spatial information of geospatial objects and find spatial patterns. For example, when applying to crime analysis, a typical output rule can be "{Within, CensusBlock79}=>{Close


probably it occurred at locations close to roads. In SAR mining, spatial objects are divided into 2 types, reference objects and task-relevant objects. Take crime analysis as an example, the crime cases are reference objects, while the roads and census blocks are task-relevant objects. Task-relevant objects are used to describe the patterns of reference objects, which are the research target of SAR mining.

For customers of Online Spatial Association Rule Mining Service (OSARMS), raw spatial data is extremely sensitive in some cases, e.g., when analyzing military and government data. Customers don't want to take the risk to share or expose raw data, even exposing encrypted raw data is not that acceptable. Furthermore, mined knowledge (e.g., rule results) is also very sensitive. Performance is not a big issue only when the analysis is executed on cloud side with high-performance computing resources. However, this requires customer to share raw data, and mined rules are exposed to service provider on cloud side. But usually in such a situation, privacy is much more important than performance. A direct and existing solution is to prepare data on customer side and mine rules on cloud side, because raw spatial data details can be hidden in data preparation step before performing rule mining. The data preparation of SAR mining refers to constructing spatial predicate transactions. SAR mining relies only on spatial relationships represented by...