alexander vodyaho & nataly zhukova — implementation of agile concepts in recommender systems...
Post on 27-Jul-2015
54 Views
Preview:
TRANSCRIPT
Implementation of Agile Concepts in Recommender Systems for Data
Processing and Analyses
Alexander Vodyaho,Nataly Zhukova
St. Petersburg Electrotechnical University “LETI”
E-mail: nazhukova@mail.ru
AIST-2015, April, 9-11, Yekaterinburg
САНКТ-ПЕТЕРБУРГСКИЙ ГОСУДАРСТВЕННЫЙ ЭЛЕКТРОТЕХНИЧЕСКИЙ УНИВЕРСИТЕТ «ЛЭТИ»
Data Processing Applications Current State
Constantly increasing number and complexity of the solved problems in the sphere of data processing
Algorithms are developed by various groups of researchers all over the world using various tools and without taking into account any standards
Constantly increasing requirements to accuracy, reliability and operativeness of data processing
Algorithms are open source and can be modified
Lack of time, financial and human resources for applications development
Specialized features of processed data require comprehensive testing of the developed algorithms
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Outline
Introduction
Problem statement
Agile concepts for IS and DPAS
Description of the developed ontologies and their relations
Description of telemetric information processing application
Case study
Conclusion
1
2
3
4
5
6
7
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Recommender Systems
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Problem statement
Requirements to the DPA RS
have to make suggestions for selecting and configuring algorithms using available information and knowledge about the subject domain of data processing and applied subject domains
have to consider information about solved problems, characteristics of analyzed objects, features of input data, parameters of the data processing systems
Required means and tools of data processing systems
means and tools for building knowledge-based descriptions of the algorithms and defining relations with the objects of the subject domain of data processing;
tools for interpreting logical rules that define conditions of algorithms efficient application and other conditions;
tools for building sequences of algorithms;
procedures for estimating the formed results;
storages of the detailed information about results of algorithms application ;
modeling tools for estimating alternatives;
Jena
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Information and Knowledge of the Subject Domain of Measurements Processing
User
Human User
Subject domain expert
Common User
Machine User
Data processing Application
External Knowledge
Base Application
Providing information and knowledge
Creation and assessment of information and
knowledge
Using result of data processing
New data processing and analyses
Retrieving new knowledge from historical data
Solving complicated task of data processing
Solving complicated specialized task
Extending available knowledge and
information
Building, Improving and estimating ontology
Application of knowledge for solving
task of data processing
Exchanging information and knowledge in standard formats
Receiving information and knowledge
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Features of the subject domain of measurements processing
• Structured binary streams, time series or separate measurements
Initial measurements
• Huge volume, bad quality, heterogeneity, distribution in time and space, non stationary behavior of time series, multiple complicated relations
Measurements features
• Formalized knowledge about measurements
Requirements to the results
• Results of data processing are used for solving tasks at the level of objects, situations and for decision making support
Consumers
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Basic agile concepts
Analyses
Design
ImplementationTesting
Evaluation
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Agile Сoncepts for DPAS
DPA RS
DPAS
DesignDevelopment
(methodological aspect)
Development (implementation
aspect)Support
First level of agile features support (industrial level)
Second level of agile features support
(research-oriented level level)
Ready methodological
solutions
Ready technological
solutions
Information Systems (IT
sphere)
Scientific prototypes
Base level of agile features support
(IT level)
Third level of agile features support
(research level)
Ready technological
solutions
Execution
Suggestions for new algorithms
Ready implementation of algorithms
Ready technological
solutions
Ready technological
solutions
Life cycle Life cycle
System Agility support
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Common and Agile Features of DPA RS
Agile features and D
PA
RS
technologies lead to new problem
sA
gile features and DP
A R
S technologies lead to new
problems
Features are based on technologies Features are based on technologies
Technologies for DPA RSTechnologies for DPA RS
Technologies for RS Technologies for DPA RS
Content-based approach
Collaborative filtering
Hybrids
Knowledge-based approach
Logical inference
Experience-based approach
Exploration analyses
DPA RS features DPA RS features
RS features Agile features
Capability to process huge amounts of data
Capability to make suggestions
Ranging capabilities
Easy integration of new methods and algorithms
Easy development of new methods and algorithms
Easy extension of data processing and analyses systems business logic
DPA RS problems DPA RS problems
Easy integration of new methods and algorithms
Easy development of new methods and algorithms
Easy extension of data processing and analyses systems business logic
Low cost of design, development and support
Short time of design, development and support
Convenient working space
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
DPA RS Information Model
Model of the applied subject domain of data processing
and analyses
Model of the applied subject domain of data processing
and analyses
DPAS dynamic information model
Information model of the environmentInformation model of the environment
Subject domain modelsSubject domain models
DPAS information modelDPAS information model
RS DPAS information modelRS DPAS information model
DPAS satic information modelDPAS satic information model
Model of the applied subject domain
Model of the applied subject domain
DPAS RS dynamic information model DPAS RS static information modelDPAS RS static information model
<<inherits>> <<inherits>>
<<inherits>> <<inherits>> <<inherits>>
<<uses>>
<<uses>>
<<inherits>>
<<uses>>
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
DPA RS Architecture
RS GUI tools GUI Processes management toolOntologies and knowledge bases editors
RS tasks manager
RS content manager
Ontologies
Service of mathematical and modeling libraries
Service for external connections management
and support
Data, information and knowledge visualization
service
Processes management and execution service
Network
Nework
Knowledge bases
Data, information and knowledge sever
Data bases File storage
Data, information and knowledge
Backend
Administrative serviceData processing and
analyses service
Data processing and analyses tools
Frontend
Inference machine
RS ontologies
RS knowledge bases
RS data, information and knowledge manager
DPAS componentsDPAS componentsDPA RS componentsDPA RS components
Research-oriented services
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Case study
The system has the aim to analyze and control structure and contents of the binary streams received from space objects.
Example of the binary steams structure
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
General procedure for the binary streams processing and analyses
Apply methods based on calculation of the frequency distribution of the streamsApply methods based on calculation of the frequency distribution of the streams
Binary streams
Compare the descriptions with the descriptions of the earlier received streamsCompare the descriptions with the descriptions of the earlier received streams
Preliminary streams descriptions
Similar streams are found?
Restore the length of the cards and words, subcommutation and supercommutation of the parameters using methods of correlation analyses
Restore the length of the cards and words, subcommutation and supercommutation of the parameters using methods of correlation analyses
The structure needs improvements?
Build and analyze the graphs that represent the structure of the streamsBuild and analyze the graphs that represent the structure of the streams
Streams descriptions
Streams descriptions
Streams descriptions
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
GUI of program complexes for processing stream structure
AIST-2015, April, 9-11, Yekaterinburg
Experimental results
Parameters of the stream
98 80 91,3 90,1
92 40 90,2 92,4
98 78 93 89,5
80 20 72,3 71,7
wL Sub SupfL
%60|| Sub
%60|| Sup
%5DP
%35SVP
%40FVP
%10FPP
%10OP
%5DP
%35SVP
%10FVP
%40FPP
%10OP
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
Contacts
Contact information:
Nataly Zhukova
5, Prof. Popova str, St. Petersburg, 197376
E-mail:nazhukova@mail.ruPhone:+7 812 346-46-41
Fax:+7 812 346-46-41
AIST-2015, April, 9-11, Yekaterinburg
SPbETU «LETI» www.eltech.ru
top related