Abstract: Aiming at the uncertainty of power business expansion project duration, an ICSO-SOM-ELM prediction model of power business expansion project duration is proposed based on self-organizing map network clustering and improved crisscross algorithm to optimize the weight threshold of extreme learning machine. Firstly, based on the project budget cost and the number of nodes, the self-organizing map network is used to secondarily cluster the data of power business expansion project, so as to preliminarily reduce the confusion of the original data set. Secondly, an improved crisscross algorithm based on the mechanism of neighborhood population crossover and mutation is proposed, which is used to optimize the weight threshold of an extreme learning machine model to obtain the optimal ELM prediction model. Finally, according to the secondary clustering data of power business expansion projects, an ICSO-ELM prediction model is used to predict the project duration. An experiment is conducted with the business expansion data of a power supply company. Results show that the proposed ICSO-SOM-ELM prediction model is better than other ones, verifying its effectiveness and providing scientific suggestions for duration planning of power supply companies’ business expansion projects.

Key words: power business expansion project, duration prediction, self-organizing map network, improved crisscross algorithm, extreme learning machine