Abstract: We proposed a forecasting model of shortterm unoccupi ed parking space by using the method of combining wavelet transform with extr eme learning machine(ELM). Firstly, the time series of the effective parking spa ce were decomposed and reconstituted by wavelet function. Secondly, ELM was used to forecast the decomposed time series respectively. Final ly, the prediction results of each neural network were combined to get the final prediction results. The prediction instance results show that the method shorte ns the training time and improves the prediction results.

Key words: wavelet transform, unoccupied parking s pace, extreme learning machine （ELM）, parking space management system