3. Introduction (1-1) :
Data mining is sorting through data to
identify patterns and establish relat-
ionships.
Data mining parameters include :
- Association;
- Sequence or path analysis;
- Classification;
- Clustering;
- Forecasting.
4. Introduction (1-2) :
4
Data is
Very
complex
So we have top 10 challenging
Problems in data mining
There is a
different
Way to extract
The information
A huge
amount
of data
Data is
power
Many
algorithms
5. - Top 10 challenging Problems in data mining (DM) :
1- Developing a Unifying Theory of Data Mining :
The developers could not have a structure that contains
the different datamining algorithms .
Knowledge
To be
verified
Types of dataset Selection criterion Unified (DM) process
Numeric
Categorical
Multimedia
Text
Akaike
information
criterion
Clustering
Classification
Association
6. - Top 10 challenging Problems in data mining (DM) :
2- Scaling Up for High Dimensional Data and High
Speed Data Streams :
The problem begins
when the data becomes
huge and complex
we need ultra-high
dimensional
classification
problems
(millions or billions
of features )
Rather than
we need
Ultra-high
speed data
stream
7. • In this problem we
want to see how to
efficiently and predict
the direction of these
data .
• In any design we must
take care of this three
master steps:
7
Practical
design
Predictor
Information
Learner
(1) QIANG YANG ,10 Challenging problems in data mining research , International Journal of Information Technology & Decision
Making , Vol. 5, No. 4 (2006) 597–604 .
- Top 10 challenging Problems in data mining (DM) :
3- Mining Sequence Data and Time Series Data :
8. • We have complex knowledge when we have mining data
from multiple relation.
• In most domains, the object of interest are not
independent of each other.
• The objects are not of a single type.
8
HTML has a tree structure
(nested tags)
Text has a list structure
(sequence of words)
Hyperlinks graph structure
(Linked pages)
Example
domains
Worldwide
Web
(1) Jarosław Stepaniuk , Rough – Granular in Knowledge Discovery and Data Mining , Volume 152 of the series , pp 99-110 .
- Top 10 challenging Problems in data mining (DM) :
4. Mining Complex Knowledge from Complex Data :
9. 5.1 : Community and social
networks :
• when we say community we must
take important topics that are
mining of social networks .
• The challenging to identify the
problem is :
It’s critical .
Distributed .
Snapshot .
9
5.2 : Mining in and for computer
networks — high-speed mining
of high-speed streams :
• This part studies how to provide
a Good algorithm are and how
to detecte an attack .
• DoS (Denial of Service) how to
detected it and how to
discriminate .
We will discuss two part in this problem:
(1) Qiang Yang, Hong Kong , 10 Challenging Problems in Data Mining Research , ICDM 2005 , pp 8.
- Top 10 challenging Problems in data mining (DM) :
5. Data Mining in a Network Setting :
10. • Need to correlate the data
seen at the various probes
(such as in a sensor
network).
• The important problem is
how to mine across
multiple heterogeneous
data sources.
• The goal is to minimize the
amount of data shipped
between the various sites,
by combining data mining
with game theory.
10
(1) Rao , Dr. S Vidyavathi , Distributed data mining and mining multi – agent data , International Journal on Computer Science and
Engineering ,Vol. 02, No. 04, 2010, 1237-1244 .
- Top 10 challenging Problems in data mining (DM) :
6. Distributed Data Mining and Mining Multi-Agent
Data :
11. 11
• The world today is “resource-driven”.
• So how we could have a best understand and
hence utilize about our environment .
• The researchers try to solve these problems :
- Bioinformatics . - Spatial data .
- Earthquakes . - Land slide .
- Biological sequence . - Cancer prediction .
() Pooja Shrivastava & Dr. Manoj Shukla , A Brief Survey On Data mining For Biological and Environmental Problems , International
Journal of Scientific & Engineering Research, Volume 4, Issue 11, November-2013 , pp630-631 .
- Top 10 challenging Problems in data mining (DM) :
7. Data Mining for Biological and Environmental
Problems :
12. Data
cleaning
• how to merge visual
interactive and
automatic (DM)
techniques together.
12
• how to perform
systematic
documentation of
data cleaning .
• Help users to avoid
mistakes in (DM).
• Create a methodology
in (DM) .
() QiangYang , 10 Challenging Problems in Data Mining Research , ICDM 2005 , pp 11 .
- Top 10 challenging Problems in data mining (DM) :
8. Data Mining Process-Related Problems :
Automate
(DM)
operations
Combine
techniques
13. 13
Knowledge integrity challenges
Knowledge integrity challenges
The challenges facing researchers
Data are being mined
Develop efficient algorithm to
compare (before & after) knowledge
contents .
Not just evaluates the knowledge integrity
But also measures to evaluate the
knowledge integrity of individual patterns.
How to mined the data with
Ensure the user’s privacy
Develop algorithms for estimating
the impact of the data.
() QIANG YANG , 10 CHALLENGING PROBLEMS IN DATA MINING RESEARCH , International Journal of Information
Technology & Decision Making Vol. 5, No. 4 (2006) , pp603.
- Top 10 challenging Problems in data mining (DM) :
9. Security, Privacy, and Data Integrity :
14. 14
Sampling
Correct the
bias
Deal with
special data
Sampling and model
building are not optimal .
Here is the problem that how
to correct the bias as we can.
Deal with unbalanced and
cost – sensitive data .
Obtaining
these costs
relied on
sampling
method .
() QIANG YANG , 10 CHALLENGING PROBLEMS IN DATA MINING RESEARCH , International Journal of Information
Technology & Decision Making Vol. 5, No. 4 (2006) , pp 603-604 .
- Top 10 challenging Problems in data mining (DM) :
10. Dealing with Non-Static, Unbalanced and Cost-
Sensitive Data:
15. Conclusions :
• The presentation highlights on the
most important 10 problems in data
mining but in concise manner .
• The order of the sequence list does
not reflect their level of important .
15
16. • We must try to work hard to overcome these problems ,
because nowadays the one who owns the information
he has the power .
16
Suggestions :
Association - looking for patterns where one event is connected to another event
Sequence or path analysis - looking for patterns where one event leads to another later event
Classification - looking for new patterns (May result in a change in the way the data is organized but that's ok)
Clustering - finding and visually documenting groups of facts not previously known
Forecasting - discovering patterns in data that can lead to reasonable predictions about the future (This area of data mining is known as predictive analytics.)
Some of the key issues that need to be addressed in the design of a practical data miner for noisy time series include:
• Information/search agents to get information: Use of wrong, too many, or too little searchcriteria;possiblyinconsistentinformationfrommanysources;semantic analysis of (meta-) information; assimilation of information into inputs to predictor agents.
• Learner/miner to modify information selection criteria: apportioning of biases to feedback; developing rules for Search Agents to collect information; developing rules for Information Agents to assimilate information.
• Predictor agents to predict trends : Incorporation of qualitative information ; multi objective optimization not in closed form .
Mining graphs
Data that are not i.i.d. (independent and identically distributed)
1-many objects are not independent of each other, and are not of a single type .
2-mine the rich structure of relations among objects .
3- E.g.: interlinked Web pages, social networks, metabolic networks in the cell .
Integration of data mining and knowledge inference .
The biggest gap: unable to relate the results of mining to the real-world decisions they affect -all they can do is hand the results back to the user.
More research on interestingness of knowledge .
First, it’s critical to have the right characterization of the notion of “community” that is to be detected.
Second, the entities/nodes involved are distributed in real-life applications, and hence distributed means of identification will be desired.
Third, a snapshot-based dataset may not be able to capture the real picture .
One result that we sure about it if we don't solve the privacy issue , data mining will become a derogatory term to the general public .
Develop algorithms for estimating the impact that certain modifications of the data have on the statistical significance of individual patterns obtainable by board classes of data mining algorithms .
Historical action in sampling and model building are not optimal , but they are not chosen randomly to . This gives the following challenging phenomenon for the data collection process .
A challenging problem is how to correct the bias as much as possible .
Many opinions of researchers who worked in this field are summarized .