Enhanced High-Dimensional Data Prediction using Additive Square Lasso-based Imputation and Deep Learning Architectures

Missing values are a common problem in high-dimensional datasets making precise analysis of data and predictive modelling extremely difficult. Conventional imputation methods frequently produce biased or less-than-ideal results because they are unable to preserve the intricate relationships and connections found in high-dimensional data. In order to tackle this issue we provide an Additive Square Lasso-based Multiple Imputation (ASLMI) technique that manages missing data efficiently while maintaining complex feature associations. Our method makes use of a dataset with 50 highly dimensional features where rigorous data preprocessing is carried out to guarantee robustness and consistency. Remarkable imputation accuracy is achieved by combining the ASLMI technique with cutting-edge machine learning models, such as Random Forest, k-Nearest Neighbors (KNN), Generative Adversarial Networks (GANs) and Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM). The R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (coefficient of determination) measure is used to assess the model’s performance. RNN-LSTM achieved a remarkable R2 score of 0.9945 while GAN, Random Forest, KNN and ASLMI all achieved R2 ratings of 1.0 indicating impute and prediction abilities. The proposed ASLMI technique demonstrates its efficacy by ensuring reliable and accurate imputation while maintaining the basis of high-dimensional data. In order to handle data that is missing in high-dimensional datasets this research helps build numerous identification algorithms that are reliable and scalable. This improves forecast accuracy and guarantees correct decisions in a variety of real-world applications including as scientific research, healthcare and finance.