Finance Industry. Finance Industry. - Particular in the stock market, predictive methods have been employed to attempt to forecast reactions to particular news and to model behaviour. Unfortunately, the widespread use of such tools in themselves changes to observed behaviour. Nevertheless, they have proven successful (although the best are understandably not publicised).
Database Systems. - In some ways a neglected area in which the results of data mining are fed back into the database system to improve aspects such as query optimisation and data distribution. They might also be used to provide soft answers to database queries.
One answer to the future directions for DMKD question will reflect those occurring in the remainder of the ICT industry. One answer to the future directions for DMKD question will reflect those occurring in the remainder of the ICT industry.You have five minutes to consult with the person next to you to come up with three broad directions of the ICT industry for each of:- Technology Push
- Consumer Pull
One answer to the future directions for DMKD will reflect those occurring in the remainder of the ICT industry: One answer to the future directions for DMKD will reflect those occurring in the remainder of the ICT industry:
… not just from the primary data source but also from linked files. … not just from the primary data source but also from linked files.Issues will include:- Scalability of algorithms,
- The use of realistic and defensible sampling techniques,
- Optimisation techniques,
- Interruptible algorithms.
The availability of more complex data such as: The availability of more complex data such as:- Temporal (and longitudinal),
- Spatial,
- Structured,
- Multimedia,
Integration Techniques.- How do we combine what we have found from the various new complex data?
Rule Semantics
As ICT turns to the tougher environments (many areas are now considered “solved” problems in terms of research) so too will DMKD apply itself to these areas. See application domains earlier. As ICT turns to the tougher environments (many areas are now considered “solved” problems in terms of research) so too will DMKD apply itself to these areas. See application domains earlier. New platforms such as PDAs and wireless devices will provide new opportunities for data capture and mining, for example,- Mining mobile web traffic (web + spatial)
- Mining over smaller on-board data and with more limited compute power.
Data mining architectures and rule semantics are something that few researchers have considered in any depth. Data mining architectures and rule semantics are something that few researchers have considered in any depth.- How do we construct a mining or visualisation routine such that it can be used and understood by other routines?
- How do we generate a generally agreed interpretation of a rule’s meaning?
- Can we, for example, plug one mining algorithm into the back of another and get useful results? (At Flinders we would contend you can).
Association Rules indicate that the occurrence of one or more items together implies the coincidental occurrence of one or more other items, according to the dataset over which the mining was done. Association Rules indicate that the occurrence of one or more items together implies the coincidental occurrence of one or more other items, according to the dataset over which the mining was done.Their syntactic form is:- Antecedent -> Consequent, (σ%, γ%)
For Example, in a set of market basket transactions- Tea, Milk -> Sugar, (10%, 65%)
- “People buy sugar because they buy Tea and Milk” is not an appropriate conclusion to make.
It applies ONLY for the dataset under evaluation.Unless the rule has σ =100% it cannot be said to apply to any given transaction.It indicates possible correlations that further research might explain or refute.
There is a difference between an association and a correlation (although the terminology is often confused). There is a difference between an association and a correlation (although the terminology is often confused).Symmetry- Correlations are generally symmetric. If A is correlated with B the B is correlated with A.
- Associations are generally asymmetric. A -> B does not imply B -> A.
Data Type- Correlations are generally associated with continuous data.
- Associations are generally associated with discrete or categorical data.
Dostları ilə paylaş: |
