Predicting fish species richness to human-introduced environmental variations has always been a vital issue for conservation management.Random forests (FR),Classification and regression trees (CART) and Self-Organizing Map (SOM) retain significant statistical techniques which can be used to predict relationships between species richness and environmental variables by repeating splitting of the data.In the present study,17 environmental variables (depth,water velocity,substrate cover (blocks,pebble,gravel,sand,silt,Marne,slab),pH,temperature,conductivity and 5 kinds of macrophytes (Myriophyllum,Ranunculus,Potamogeton,Vallisneria and Jussiaerepens),and 17 fish species were collected from 325 sampling sites (65 X 5 stations: Rejet,HR_RD,HR_RG,LAM and LAS) around the nuclear power station which was located in the downstream of Garonne River of France.Rejet was an area directly influenced by the rejection of the Nuclear Power Station; HR_RD and HR_RG were the right and left sides that were out of influence area of rejection respectively; LAM and LAS were two sampling stations which could be influenced by the rejection located at a distance of 1.6 km and 5 km respectively.In the present study,we aimed to first predict the 5 stations with environmental variables and also to infer the effect of rejection of the nuclear power station on fish species richness with machine leaning approaches.Index Gini and Accuracy inferred three environmental variables conductivity,temperature and pH were the most important contributions in distinguishing the 5 stations using RF.Similar results were indicated using CART,in which HR_RG differed due to the lower pH values (pH<8.315),and with conductivity ≥297.5,station HR_RD and Rejet were separated from station LAM and LAS,sequent,temperature separated the two stations in the two cluster.Supervised SOM inferred the response of fish species richness to the influence of this nuclear power station.SOM showed that only 2 fish species were observed within Rejet (direct pollution zone),and as the decrease of influence of the rejection,we observed improvement of specific richness with 3 species at LAM and 6 species at LAS.Out of the rejection influence,more fish species were indicated within HR_RD and HR_RG,with 4 and 7 respectively.Therefore,we proposed that fish species richness was affected by the nuclear power station and management of water pollution should be taken in the area.