Uncertainty Reduction of Unlabeled Features in Landslide Inventory Using Machine Learning t-SNE Clustering and Data Mining Apriori Association Rule Algorithms

A landslide inventory, after an intense rainfall event in 1998, Southwestern Korea, was collected by digitizing aerial photographs. This left high uncertainty the inventoried features to be verified ground truths. To reduce uncertainty, photographs were reexamined, supported time slider Google Earth. We observed 77 deformed slopes, which similar shape and texture, landslides. then sought label formations based on their spatial relationship with surrounding conditions. three-phase methodology developed. First, inventory of landslide, no vulnerable unlabeled analyzed cluster patterns, dimension reduced using t-distributed stochastic neighbor embedding (t-SNE). Second, Apriori algorithm, association rule mining, used identify common relations antecedent factors (derived from topographic landcover maps) that are linked areas features. Third, findings validated Landsat TM (Thematic mapper) ETM+(Enhanced thematic images acquired before original inventory. Current research offers practical economical solutions (reduced reliance paid remote sensing sensors field survey) labeling classification missing or outdated attributed information.