### Table 2.2: Computation times from simulation experiments with naive Bayes models

2007

### Table 2.2: Computation times from simulation experiments with naive Bayes models

2007

### Table 2.2: Computation times from simulation experiments with naive Bayes models

2007

### Table 2.2: Computation times from simulation experiments with naive Bayes models

2007

### Table 2.2: Computation times from simulation experiments with naive Bayes models

2007

### Table 2: Sfi results for naive Bayes and OB1 incorporating naive Bayes models using hold-one-out weighting on discretised UCI datasets

1998

"... In PAGE 13: ...4 the probability of a data value v if one in- stance of it is held out is P (vjD ? fvg) = cv c+n?2: Each symbol v is held out cvtimes so the total contribution of that symbol to the probability is cv c+n?2 cv . Thus P (D) = Y v cv c + n ? 2 cv = (c + n ? 2)c Y v ccv v Table2 shows the Sfi for OB1 when including naive Bayes models in the tree and using hold-one-out weighting. Experimental conditions are the same as for the previous experiments.... ..."

Cited by 1

### Table 1. Comparison of different classification methods: decision tree models, nearest neighbour approaches, naive Bayes model and multi-layer perceptron. Sign + means that the model supports the property, - that it does not.

"... In PAGE 3: ... However, we have tried to evaluate and compare the most common classification methods (decision trees, nearest neighbour methods, naive Bayes model and multi-layer perceptrons) according to the general requirements of context-aware systems. ( Table1 ). The first criteria deal with efficiency of reasoning (i.... ..."

Cited by 1

### Table 5: Percentage correct for machine learning schemes, on discretised UCI datasets, relative to OB1 using hold-one-out weighting, naive Bayes models, with tree depth bound to 3 attributes

1998

"... In PAGE 15: ... OB1 settings are currently considered default: incorporating hold-one-out weighting, including naive Bayes models, and tree depth bound to three attributes. Table5 shows the percent correct results. Where a scheme performs signi cantly better than OB1 this is post xed with , and where a scheme performs signi cantly worse than OB1 this is post xed with .... In PAGE 16: ...Table5 are impressive, with 8 cases where a scheme performs signi cantly better than OB1, and 59 cases where schemes perform signi cantly worse than OB1. As OB1 is still a work in progress, we intend to perform more detailed comparisons with other schemes, including boosted algorithms.... ..."

Cited by 1

### Table 3. Applying the Naive Bayes Word Pair Model to unambiguously marked data, 10-fold cross-validation

2005

"... In PAGE 25: ...). The results for the Naive Bayes word pair model are shown in Table3 . The overall accuracy obtained by this model is fairly low at 42.... ..."

Cited by 2

### Table 2. Comparison of our Bayes net and naive Bayes models. Results for models constructed using all features, individual feature groups and those features derivable from only gene coordinates and sequence (gene spacing, operon length and codon usage)are presented. The first column indicates the features used to construct the model. The second and third columns show the areas under the ROC curves (AUC) of our Bayes net model and a naive Bayes model respectively. The fourth column shows the p-values from a test of the null hypothesis that the AUCs of the two models are the same

"... In PAGE 7: ...ot have the rich data sources that are available for E.coli. Furthermore, as long as gene boundaries and sequence are known, the operon length and codon usage features can be easily obtained as well. In fact, experiments on the Bayes net model with the gene spacing, operon length and codon usage features yield performance similar to the full model (last line in Table2 ). Of course, making predictions presupposes the existence of a model trained from a training set of known operons and non-operons.... In PAGE 7: ... In addition we run experiments with models containing the features derivable from only gene coordinates and sequence data (the gene spacing, operon length and codon usage features). Table2 shows the area under the ROC curves of Bayes net and naive Bayes models using the different feature groups. Also shown are the p-values from a test of equal areas.... ..."