Anomaly detection and data imputation within time series
Data mining(1)
1. Explored Research on Data Preprocessing and Mining
Technology for Clinical Data Applications
GROUP MEMBERS
1. S.M.ZAHIDUL ISLAM(152-15-5925)
2. MD.HABIBUR RAHMAN(152-15-6040)
3. AKTERUZZAMAN(152-15-5671)
4. ABU MOHAMMAD MUSA(152-15-5682)
2. Overview
Abstract
INTRODUCTION
DATA PREPROCESSING
DATA MINING
CLINICAL MEDICAL DATA
BIOCHEMICAL INDICATORS
DATA PREPROCESSING AND NETWORK MODELING
CONCLUSION AND DISCUSSION
4. INTRODUCTION
People felt
uncomfortable for they
would go to hospital.
Development area
(data mining technology,
information science,
improvement of HIS)
Hidden essence
relationships
Combining information
science with clinical
medical data.
5. Data Processing
Design defects of medical equipment
Subjective factors
Experimental environment force majeure
Noise
Redundancy
Mistakes
8. CHARACTERISTICS OF CLINICAL
MEDICAL DATA
Clinical
Medical
Data
1.
Privacy.
2.
Diversity.
3.
Complexit
y.
4.
Incomplet
eness.
5.
Redundan
cy.
6.
Heteroge
neity.
10. Data preprocessing and network modeling
1. Data Preprocessing: It is quality assurance of data source, which can
greatly improve the accuracy and performance of dataminin.
2. Data Mining: It is an information processing technology .
knowledge from database
Artificial intelligence
statistics and pattern recognition
Visualization technology and parallel computing ect.
11. Flow chart
Integration &
Association
Cleaning
Filling & de-
noising
Normalization
Input of Network
Bio chemical indicator
Data
selection
Preprocessin
g
Clustering
&mining
Making use of
biological
computational
methods, statistical
theory, pattern
recognition and
artificial intelligence in
data mining, at present
SOM network is
adopted .
Analysis &
explanation
Feedback
Adjustment
Information Obtainment of
Health
Assessment, Trend Analysis
and Disease Intervention
12. Conclusion & Future work
SOM does not need training set after adjusting learning function of weight
and threshold.
Base for future study on early warning and intervention of health status.
Provide greatest convenience and objective benefit for assisting doctor
diagnosis and medical decision management
Introduce manifold learning of geo-information science into data mining
on clinical data
Problem: how to determine slow changing part around peak is its sub-
classification or a new classification .