Adaptive neural network system to build environmental prediction and control by their typing biometrics

Abstract

© 2020. All Rights Reserved. Energy conservation, environmental protection, and intelligence are topics of interest in intelligent buildings. However, the energy requirement of various electrical equipment in smart buildings increases energy consumption. This paper presents a neural network-based prediction and control system for the regulation of building environmental parameters and discusses the problem of recognizing users by their typing biometrics. The expediency of its solution based on the training of a neural network is noted. The biometric authentication algorithm is described; the principles of this algorithm functioning, as well as the developed technology of biometric users authentication, are considered. This problem solution automating required the collection and preparation of initial data for analysis, neural network model constructing, as well as the research conducting and the accuracy of biometric user authentication based on the constructed models assessing. To prepare the initial data and form the training sample for the neural network training, a dataset consisting of 500 users typing biometrics templates, containing a username and a passphrase was created. A neural network model was constructed on the basis of the prepared data. The result of the calculated values ("Legal user" or "Illegal user") was used as an output feature. The research has shown, that the amount of the 1st type errors (the number of illegal users classified as legal) was 0%, and the value of the 2nd type errors (the number of legal users classified as illegal) was 3.3% The percentage of correctly classified users based on the trained neural network was 96.7. Thus, the developed neural network system can be effectively applied for biometric user’s authentication by using typing biometrics.