Kullanicilarin nöro fiizksel durumlarini anlayarak insan odakli riskleri azaltmak

Today, smart devices are capable to collect their users' data using a variety of different sensors. Devices are usually online and they share their data using internet. This device to device communication is called as IoT (Internet of Things). IoT backbones store data in Data Centres (DC). Stored data contains valuable information about the users. Analysing the data gives detailed information about the users. In this thesis, a small model of this above IoT structure is simulated. Data is collected and emotion analysis of the users is improved by enriching the physical characteristics of users such as sleep, heartbeat, mobility, etc. with neuro physical parameters such as keystroke patterns, and keystroke error count measurement. The contribution of this work to the literature is the combine usage of physical and neuro physical parameters. Keystroke patterns, and keystroke error count measurement are studied as neuro physical parameters while sleep quality, energy, mobility/movement, and heart pulse are analysed as physical parameters. Another novelty is that the classification task is performed by both conventional(classical) machine learning algorithms and deep learning models to analyse the emotion of the users. For this purpose, feedforward neural network (FFNN), convolutional neural network (CNN), recurrent neural network (RNN), and long short-term neural network are employed as deep learning methodologies while multinomial naive Bayes (MNB), support vector regression (SVR), decision tree (DT), random forest (RF), and decision integration strategy (DIS) are evaluated as conventional machine learning algorithms. To the best of our knowledge, this is the very first attempt to analyse the neuro physical conditions of the users by evaluating deep learning models for sensitivity analysis and enriching physical characteristics with neuro physical parameters. The dataset is collected with the usage of smart devices and sensors from the users during one-year time period. Experimental results demonstrate that the utilization of deep learning methodologies and the combination of both physical and neuro physical parameters enhance the classification success of the system to interpret the sensitivity of the users. A wide range of comparative and extensive experiments show that the proposed model exhibits noteworthy results compared to the state-of-art studies.