The solution of the "cold start problem" in e-Learning

Problem Statement. Among the main problems of e-Learning one is the "cold start problem". A learning environment cannot provide information on a relevant difficulty of the content due to the lack of information about a learner. The unsolved problem entails a reduction in the effectiveness of the learning process as overly difficult or, on the contrary, too easy content leads to a loss of learning motivation, frustration and stress among students. The study is aimed at searching the solution of the "cold start problem". Research Questions. Which model of detecting the starting difficulty of the content will be universal and will produce a stable forecast for different samples? Purpose of the Study. Developing the model for detecting the optimal starting difficulty of the content, including searching predictors meets the criteria of the universality. Testing hypotheses about the sustainability of the model for samples with different levels of preparedness. Research Methods. The model developing based on Logistic regression and Item Response Theory. Testing hypotheses about sustainability based on hybrid simulation. This simulation type used real predictors and generated data sets (parameters of the content difficulty) simultaneously. The data sets have made by Monte-Carlo software. Simulation has replicated several times for checking the sustainability criteria. Findings. Verbal and numerical intelligence parameters are potent and universal learning efficiency predictors. The detected starting level of the content difficulty is close to simulated learners' level of knowledge (preparedness). The model is statistical significant and sustainable in cases of samples with similar or different levels of knowledge (preparedness).