Data Mining: Practical Machine Learning Tools and Techniques, Second Edition

tions, as described in Section 4.2; alternatively they can be treated as an addi- tional value of the attribute, to be modeled a ...

6.6 CLUSTERING 269 mixture model, problems will occur whenever any of the normal distributions becomes so narrow that it is cent ...

for the numeric attributes. This involves an additional, outer level of search. For example, it initially evaluates the log-like ...

6.7 BAYESIAN NETWORKS 271 the X-means algorithm of Moore and Pelleg (2000). However, instead of the MDL principle they use a pro ...

Does this mean that we have to accept our fate and live with these shortcom- ings? No! There is a statistically based alternativ ...

6.7 BAYESIAN NETWORKS 273 outlook;the right gives a probability for each value oftemperature. For example, the first number (0.1 ...

274 CHAPTER 6| IMPLEMENTATIONS: REAL MACHINE LEARNING SCHEMES play play yes .633 no .367 outlook play yes no outlook sunny .238 ...

6.7 BAYESIAN NETWORKS 275 For example, consider an instance with values outlook =rainy, temperature = cool, humidity =high,and w ...

which is exactly the multiplication rule that we applied previously. The two Bayesian networks in Figure 6.20 and Figure 6.21 ar ...

6.7 BAYESIAN NETWORKS 277 of the third row of the humiditynode’s table in Figure 6.21), observe from Table 1.2 (page 11) that th ...

given network. It turns out that this can be implemented very efficiently by changing the method for calculating the conditional ...

6.7 BAYESIAN NETWORKS 279 values for the nodes in its Markov blanket. Hence, if a node is absent from the class attribute’s Mark ...

Bayesian multinet. To obtain a prediction for a particular class value, take the corresponding network’s probability and multipl ...

6.7 BAYESIAN NETWORKS 281 what the tree in Figure 6.22(b) does, because it contains only 8 counts. There is, for example, no bra ...

For example, Figure 6.22(b) contains no expansion for windy =falsefrom the root node because with eight instances it is the most ...

6.7 BAYESIAN NETWORKS 283 ers to these instances rather than a list of pointers to other nodes. This makes the trees smaller and ...

...

In the previous chapter we examined a vast array of machine learning methods: decision trees, decision rules, linear models, ins ...

is not necessarily the one that performs best on the training data. We have repeatedly cautioned about the problem of overfittin ...

attributes for learning schemes to handle, and some of them—perhaps the over- whelming majority—are clearly irrelevant or redund ...