Abstract | ||
---|---|---|
Naive Bayesian classifier is a commonly used classification algorithm, which has the advantages of high classification efficiency and low cost. However, in practical application, the assumption of class conditional independence reduces the accuracy of algorithm classification. In order to solve the problem, the flower pollination algorithm (FPA) is adopted to optimize Naive Bayes classifier, and the Naive Bayesian classifier algorithm based on improved flower pollination algorithm (NBC-IFPA) is proposed. Firstly, the blacklist mechanism is introduced to make the FPA jump out of the local optimal solution. Secondly, the random perturbation term is introduced to increase the diversity of the population and improve the searching ability of FPA. Finally, the improved FPA is used to search for the global optimal attribute weights and use them into the weighted naive Bayesian model for classification. The simulation results show that the NBC-IFPA algorithm has higher classification accuracy. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1109/WISA.2017.28 | 2017 14th Web Information Systems and Applications Conference (WISA) |
Keywords | Field | DocType |
naive bayes,flower pollination algorithm,blacklist mechanism,random perturbation,attribute weighting | Convergence (routing),Population,Naive Bayes classifier,Conditional independence,Blacklist,Algorithm,Blacklisting,Statistical classification,Jump | Conference |
ISBN | Citations | PageRank |
978-1-5386-4807-0 | 0 | 0.34 |
References | Authors | |
3 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Zi-Qing Wang | 1 | 0 | 0.34 |
Mingxin Zhang | 2 | 0 | 0.68 |
Jin-Long Zheng | 3 | 0 | 0.34 |
Guo-Hai Zhang | 4 | 0 | 0.34 |