Dear Editor: We wish to submit an original article for publication in IEEE ACCESS, titled “MetricSifter: Feature Reduction of Multivariate Time Series Data for Efficient Fault Localization in Cloud Applications”. In this paper, we propose a feature reduction framework designed to accurately identify anomalous monitoring metrics caused by faults in cloud applications to address the challenges of fault localization with multivariate time series data. This manuscript has not been published or presented elsewhere in part or entirety and is not under consideration by another journal. We have read and understood your journal’s policies, and we believe that neither the manuscript nor the study violates any of these. There are no conflicts of interest to declare. Thank you for your consideration. I look forward to hearing from you. Sincerely, Yuuki TSUBOUCHI Hirofumi TSURUTA Our proposed feature reduction framework and its experimental results show that our reduction technique based the failure time localization algorithm could greatly contribute to the efficiency of software engineering for cloud applications, and we believe that this paper will be a match for regular submission. This article fits software engineering, cloud computing, and data mining categories. The contributions of our study are summarized as follows: (1) For the first time in the literature, we quantitatively evaluate different feature reduction methods on their own. To quantitatively evaluate these methods, we formulate feature reduction as the task of classifying whether monitoring metrics are related to a failure or not. (2) We propose a feature reduction framework called MetricSifter that focuses on the proximity of change point times during a failure across monitoring metrics. Our algorithm locates a failure time segment with the highest density of the change point times. (3) We conducted the simulation and empirical experiments. The results show that MetricSifter outperforms several baselines in terms of the contribution to the various fault localization methods. The experiments demonstrate the effectiveness of feature reduction for fault localization methods. Furthermore, MetricSifter has the best overall impact on the accuracy and the time efficiency of fault localization.