Gözde Bakırlı, Derya Birant, and Alp Kut
Neural Network, Classification, Data Mining, Incremental Mining, Sensitivity Analysis, Performance Analysis
Typically, data mining is applied on data warehouses which need to be updated frequently. In that situation, trained part of the data warehouse has to be retrained after each update operation, repetitively. This study proposes a new incremental neural network algorithm (INNA) to avoid this repetitive operation and to decrease time needed for training. Experimental results show that our incremental neural network algorithm decreases training time, considerably. This study also includes the sensitivity analysis of the neural network parameters and comparison of the neural network algorithm types by using specific datasets. A new tool, Neural Network Modeller (NNM) is designed and developed for this study, and all analyses are applied with this modeller tool.
Important Links:
Go Back
