This thesis describes the research of an aspect of human-computer interaction (HCI) in the domain of mobile applications. The main motivation for this work was to develop a technique that would support the process of development of mobile interactive systems by utilizing predictive evaluation of interaction efficiency. In this respect a novel model for predictive evaluation of interaction efficiency has been developed, which specifically targets mobile applications intended to be run on present-day touchscreen devices. The proposed model identifies basic interaction elements typical for touchenabled screens, such as single- and multi-finger touch gestures, as well as tilt-based gestures for devices equipped with motion sensors. An empirical research with carefully prepared experimental setups has been carried out in order to provide execution times for the model operators, and to investigate the validity of the model. In comparison to available related work, the proposed model provides the best abstraction of touchscreen interaction, along with task execution time predictions that fairly correspond to efficiency achieved when using real mobile applications. The thesis itself consists of the following sections. In the introductory Section 1 basic concepts from the field of HCI are introduced, motivating the need for predictive evaluation models for the set of mobile touchscreen-based devices. Section 2 Interaction Efficiency as an Attribute of Mobile Applications Usability considers usability of mobile applications in general, and focuses on interaction efficiency which represents one of its main attributes. It is shown that traditional research methods known from usability engineering still prevail in contemporary studies tackling mobile applications usage. Section 3 Basic Concepts of Cognitive Modeling introduces the basic model for cognitive analysis of human activity on an interactive system interface. Central principles for translating descriptions of human performance into quantitative indicators are described, which are based on perceptual, cognitive, and motor aspects of human behavior. In Section 4 Interaction Modeling Based on Cognitive Task Analysis, special emphasis is given on task description, as it often represents the essential measureable entity when analyzing interaction efficiency. Several GOMS-based techniques for interaction modeling are presented, all of which use both motor and cognitive operations for defining task execution methods. Predicting Interaction Efficiency with the Keystroke-Level Model is described in Section 5. The basic KLM model is introduced, which originally deals with interactive applications in the desktop domain, along with a thorough study of its principles and usage. The concept of task execution time prediction is elaborated, based on elementary model operators representing motor activities, mental preparation, and system response. Section 6 Enhancements of the Model for Predicting Interaction Efficiency describes alternative approaches for interaction modeling and efficiency prediction, along with extended versions of the basic KLM model. Although there are adaptations of this model which target the domain of mobile applications, the lack of an appropriate model for interaction with mobile touchscreen devices is confirmed. The next three sections are devoted to the introduction, elaboration, and validation of a novel model for predictive evaluation of interaction efficiency. In Section 7 New Model for Predicting Interaction Efficiency on Touchscreens, interaction elements which can be found in present-day touchscreen devices are identified, and a new operator set is accordingly devised. This model contains simple touch modalities, dragging gestures, basic multi-touch gestures, as well as tilt-based motion gestures. All operators are thoroughly explained with respect to different contexts of use, here including diverse interaction styles and various form factors (screen sizes) of mobile devices. Altogether five interaction styles were identified in the study, along with five mobile device classes, here including an emulator platform run on a touchscreen monitor. Guidelines for experimental settings are provided in order to determine operator execution times, as well as simple heuristic rules for coding mental activities into task execution methods. Section 8 Determining Execution Times of the Model Operators describes the empirical research carried out in order to find both operators' execution times and the originally introduced factor μ used for execution time parameterization according to interaction style and screen size used. For this purpose a special testing application running under the Android OS has been developed, which is able to gather all relevant indicators when executing a particular operator. A special strategy for statistical data analysis is introduced and explained, which helped in differentiating operators' execution times among different contexts of use. At the end of this section, the full version of the proposed model is systematically presented. Section 9 Model Validation provides the description of the experimental apparatus used for validation of the model. Model predictions of task execution times were compared with empirical results obtained from an experiment including the measurement of real-life mobile applications usage. The accuracy of the model estimation was expressed by means of the RMSPE (Root Mean Square Percentage Error) metric, which showed to be 10.6%. In addition, the comparative analysis of mobile applications' task execution times among different contexts of use has been performed. The final Section 10 comments on the research, providing a critical overview, future work suggestions, and an outline of research contributions thus achieved. Regarding the above, the research contributions of this dissertation thesis consist of: (i) a new model for predictive evaluation of interaction efficiency for touchscreens, based on the definition of a set of keystroke-level operators and the determination of the respective execution times; the execution times are further parameterized with respect to touchscreen device classes and specific interaction styles, (ii) the experimental validation of the proposed model through precision assessment of the provided task execution time estimates; the achieved prediction accuracy (RMSPE 10.6%) shows that the model provides a good fit for real-life scenarios, (iii) the usage of real mobile applications has been comparatively analyzed among different mobile device classes (and the emulator platform) and diverse interaction styles; interaction efficiency was assessed through empirically measured task execution times.