The spatial and temporal distribution patterns of benthic macroinvertebrates were studied in an intermittent river in Algeria (Chelif wadi, North Africa), by using the self-organizing map (SOM), an unsupervised artificial neural network. The samples were collected monthly at 8 sampling sites (630 to 20 m above sea level) and community variation was analysed in space and time. Overall, the study sites showed a very poor macroinvertebrate fauna: more than 60% of samples contained less than 11 species, and 99% had less than 30 species. Furthermore, most species displayed very low abundance: 66% of the species were represented by less than 20 individuals (abundance). Among the identified taxa, Chironomidae was the dominant taxon at all sampling sites except at the most upstream site (630 m a.s.l.) where it was replaced by Coleoptera. Concerning monthly changes, the species richness was very low in August and October. Through the learning process of the SOM, samples were classified into four clusters by the SOM, and the classification was mainly related to the location of the sampling sites. Benthic macroinvertebrates were divided into four classes, which revealed the influence of pollution on their longitudinal distribution in this stream. According to the distribution gradients of the environmental variables on the SOM map, their influence on the classification of the sampling sites could be assessed effectively.