Impact of K-NN imputation Technique on Performance of Deep Learning based DFL Algorithm

Deep Learning based device free localization (DFL) technique is gaining momentum due to the contribution of deep model in target location scenario. The performance of the model is not only depending on parameters used but also on the dataset used. If the dataset is an incomplete one the performance of the model is affected unless a suitable preprocessing technique that handles the missing data is chosen. Our work proposes a novel K-NN imputation technique during preprocessing of the data for handling missing data problem and assessing the performance of deep learning based DFL model in target location estimation. The performance of the deep learning model has improved further with K-NN imputation technique. The performance measures used for assessing the impact of K-NN imputation based pre-processing are average accuracy, categorical accuracy and precision.