Single-cell RNA sequencing or scRNA-seq is a prominent clinical next-generation sequencing technology that enables researchers to analyze gene expression over individual cells. One of the serious issues in scRNA-Seq data is the presence of biological zeros and technical zeros, which generate noise and introduce sparsity, making it challenging to uncover molecular insights related to diseases. Imputation techniques face significant challenges when applied to high-dimensional scRNA-seq datasets, especially when dealing with a limited number of cells over many genes. In such scenarios, the variability in gene expression data is heightened, making it difficult for imputation methods to differentiate between genuine biological and technological zeros. In scRNA-seq, dropout events occur when gene expression is not detected despite the gene being actively expressed. This issue is compounded with small cell numbers, as it becomes more challenging to accurately impute missing values. Machine learning-based imputation methods, in particular, are prone to overfitting the data, which can result in imputed values that do not generalize well to new or unseen data. In this study, K-Nearest Neighbor (KNN) and Adaptively thresholded Low-Rank Approximation (ALRA) imputation techniques were evaluated to recover the technological zero values of the scRNA-seq dataset. The performance of imputation techniques on imputed data is assessed using Root Mean Square Error (RMSE) and machine learning-based clustering methods. From the results of the two imputation techniques, it has been observed that ALRA performs better than KNN with an RMSE of 2.83.