The appropriate categorization of tumors from a vast quantity of Gene Expression Data (GED) is one of the most difficult processes in clinical diagnosis. To combat this challenge, a Weighted Consensus of Lion Optimized K-means Ensemble with Peak Density Clustering (WECLO K-means-PDC) algorithm has been developed that calculates the Symmetric Neighborhood (SN) correlation among Data Points (DPs) using the lion optimization. An SN Graph (SNG) was created to select the number of Cluster Centroids (ClustCenter) at every iteration of clustering. But, it was not suitable if the dataset was sparse, as well as the constant density threshold may influence the distinguishing DPs within the cluster boundaries. In this paper, an Adaptive Local resultant Force neighborhood PDC for WECR K-means (WECLO K-means-ALFPDC) algorithm is proposed, which considers additional measures to determine the symmetric neighborhood correlation among the DPs when the dataset is sparse. The major goal is to consider the distinct variances between the sizes and orientations of the DPs nearer to the ClustCenters and edges. To find such variations, two novel local measures called Centrality (CR) and Coordination (CO) are introduced instead of SNG to select the ClustCenters and obtain more precise clustering for classifying cancer from genomic data. Finally, the test results show that the WECLO K-means-ALFPDC algorithm attains 88.7%, 89.1%, 88.42%, 88.38% and 89.04% accuracy on leukemia, lymphoma, prostate cancer, SRBCT and breast cancer databases, respectively compared to the WECLO K-means-PDC algorithm.