#TITLE_ALTERNATIVE#
Knowledge discovery in data (KDD) is a process of discovering interesting knowledge from large amounts of data, and data mining is one of its process. Data mining is a technology to extract useful pattern. It has two methods: predictive and descriptive. At predictive method, Bayesian network (BN) is...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/8301 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Knowledge discovery in data (KDD) is a process of discovering interesting knowledge from large amounts of data, and data mining is one of its process. Data mining is a technology to extract useful pattern. It has two methods: predictive and descriptive. At predictive method, Bayesian network (BN) is one of classification technics which is constructed using d-separation concept. Some of its algorithms need node ordering (NO). At descriptive method, sequential pattern (SP) is one of association technics, which represents time sequence of events. This research is aimed to discover relevancy between NO and SP, especially the usage of SP as NO information to algorithm of learning BN from data. NO written by notation (n1,n2,n3,n4,...ni) and SP by notation <s1,s2,s3,s4,...sj> have similarities in notation and the order of node appereance. Some algorithms for learning BN from data can use NO to find proper cut which reduces total number of conditional independency test (CI test). SP uses frequent itemset to discover the sequence. This research is focused on exploring characteristics of NO and SP, shows that NO implies not only temporal information but also causal information, while at the other hand SP represents only temporal information. Therefore, SP can be utilized for constructing NO. This can be achieved by adding some causality tests. This thesis concludes that SP represents temporal information can be utilized for constructing NO by adding some causality tests. SP that represents temporal information is subset of NO that represents both temporal information and causality. Future research can be done on dynamic Bayesian network area that represents temporal information. <br />
|
|---|