| Data Mining |
Data mining is a
general term used for technologies that analyze data to find patterns and
relationships. Data mining techniques can be used where expert knowledge is
limited and provide insight into the processes or events being studied. Although
data mining has been very successful in various applications, and has been promoted
extensively, its usefulness is affected by ambiguous or noisy data or by changing
circumstances where past data no longer represents the system. |
| Expert Systems |
Use rules and
problem-solving methodologies acquired from a human expert. They are very useful for
distributing expert knowledge and decision-making skills to other parts of the
organization but can be costly and are only suitable for narrowly defined domains. |
| Neural Networks |
A mathematical
system consisting of a large number of interconnected "neurons" that
determine the value of outputs based on the values of inputs. These systems are
often used for predictive modelling and forecasting where there are plenty of good quality
data for "training" but the models are difficult to understand. |
| Fuzzy Logic |
Solve problems
based on rules containing ambiguous ("fuzzy") linguistic variables. Used
where knowledge is limited and reasoning is not precise. Membership functions must
be defined for each variable, which can be difficult and time-consuming. |
| Genetic Algorithms |
Use biological
concepts of evolution and natural selection to find optimal solutions by
"mating" the most fit solutions in a population. Work well for complex
non-linear optimization problems but for very large problems may take a long time to
converge to an acceptable solution. |
| Rough Sets |
A data mining
technique that varies the data precision to make patterns more visible while maintaining
discernibility. Can be used with any type of data (numeric or categorical) to
represent knowledge as easy to understand logical rules. Because decisions are
categorical, it cannot be used in situations where numerical prediction is required. |
| Approximate Reasoning |
Make decisions from rules when
there is no exact match by measuring "similarity", "dissimilarity" and
"distance" between the rule and the current data. Used to overcome
difficulties in modelling due to ambiguous data or changing systems. Some testing
and tweaking of approximate reasoning parameters is needed to achieve optimzal results.
|
| Intelligent Agents |
Small computer programs that
have enough reasoning ability to perform one well-defined task. |
| Hybrid Systems |
Use two or more technologies in
combination. Proper choice of technologies gives all the advantages of each
technique while offsetting the limitations. |
Machine Intelligence
Technologies