Western Yunnan Province covers an area of about 220 000 km2 in Southwestern China, and is located at the boundary between the Indian Plate and the Yangtze Plate. It is an important part of the Eastern Tethyan ore-forming belt. The region contains numerous base and precious metal deposits, including a few super-large ore deposits, such as the Pulang porphyry Cu deposit, the Beiya and the Laowangzhai hydrothermal Au deposits, and the Jinding MVT Pb-Zn deposit. It is difficult to decompose anomalies related to mineralization from the geochemical data with multi-patterns using conventional statistical methods, such as multi-variate statistical analysis. In this study, the spectrum-area (S-A) fractal method based on a multi-fractal model is adopted to separate anomalies from Cu, Zn and Au stream sediment data. The model results are as follows. Using the S-A method, both Cu and Au local anomalies can be effectively decomposed from the high geochemical backgrounds that arise from the Permian basalt, while the local geochemical anomalies of Zn are identified along a North-Northeast trend. The distribution patterns of local Cu, Au and Zn geochemical anomalies are useful for indicating areas where new deposits may exist. Regional and local anomaly maps of Cu were depicted based on different thresholds obtained from the log–log plot. We showed that multi-fractal methods effectively and efficiently decompose geochemical data with complicated patterns into different, meaningful indictor components compared to conventional statistical methods.