In　machine　learning-driven　landslide　prediction　models,　data　quality　directly　determines　the　accuracy　of　landslide　predictions.　Currently,　data　preprocessing　methods　mainly　involve　fitting　the　data　using　moving　averages　or　least　squares　methods.　While　these　methods　are　effective　and　convenient　for　processing　global　data,　they　face　limitations　when　dealing　with　local　data　during　heavy　rainfall　events.　The　sharp　increase　in　cumulative　landslide　displacement　during　significant　rainfall　makes　it　challenging　for　overall　data　preprocessing　methods　to　effectively　handle　local　data,　leading　to　potential　data　loss　and　compromising　data　quality.　To　address　this　limitation,　this　paper　proposes　a　segmented　data　processing　approach.　During　periods　of　heavy　rainfall,　cubic　spline　interpolation　is　employed　to　interpolate　missing　data,　providing　a　better　representation　of　data　variability.　For　smaller　rainfall　amounts,　linear　interpolation　is　utilized.　Finally,　the　LOESS　smoothing　method　is　applied　to　further　enhance　data　quality.　These　meticulous　data　processing　steps　aim　to　accurately　capture　the　changing　trends　in　rainfall-induced　landslides.　By　providing　more　reliable　inputs　to　machine　learning　prediction　models,　this　approach　seeks　to　improve　the　accuracy　and　reliability　of　landslide　predictions.　©　2024　IEEE.