Research Article
When Collective Knowledge Meets Crowd Knowledge in a Smart City: A Prediction Method Combining Open Data Keyword Analysis and Case-Based Reasoning
Table 9
Conventional algorithms considered in the experiment for performance comparison.
| | Algorithms | Options (for Weka) | Value |
| | SMO | The complexity constant C | 1 | | Number of folds | 5 | | Kernel type | PolyKernel | | The epsilon for round-off error | 1.0E−12 | | Tolerance | 0.0010 |
| | BayesNet | Search algorithm | K2 | | Maximum number of parents | 2 | | Score type | Entropy | | Estimate algorithm | Simpler estimator | | Estimate algorithm option | 1.0 |
| | IBk | Number of nearest neighbors (k) | 1 | | Nearest neighbor search algorithm | LinearNNSearch |
| | Logistic | The ridge in the log-likelihood | 1.0E−8 (default) | | The maximum number of iterations | −1 |
| | C4.5 | Pruned/unpruned decision tree | Using unpruned tree | | Minimum number of instances per leaf | 2 | | Seed for random data shuffling | 1 |
| | Ripper | Number of folds for REP | 3 | | Minimal weights of instances within a split | 2.0 | | Whether not to use pruning | Using pruning |
| | NRBNF | Number of clusters to generate | 2 | | Maximum number of iterations for the logistic regression | −1 | | Minimum standard deviation for the cluster | 0.1 |
|
|
