The adoption and use of various eLearning systems to enhance the quality of education in secondary schools in Tanzania is becoming common. However, there is little evidence to suggest that students actually use them. Existing studies tend to focus on investigating students' attitude towards using these systems through surveys. Nonetheless, data from surveys is normally subject to the possibility of distortion, low reliability, and rarely indicate the causal effects. This study adopted WEKA and KEEL as data mining tools to analyze students' usage patterns and trends using 68,827 individual records from the log file of the Halostudy system implemented in secondary schools in Tanzania. The study found that the system usage is moderate and in decline. There is also variability in the usage of multimedia elements with biology having the highest number while mathematics has the lowest. Students from Dar es Salaam, Mwanza, and Arusha, in that order, had the highest system usage with the lowest being from the peripheral regions. The possible challenges limiting system usage are discussed. These findings show that data mining tools can be used to indicate usage patterns of systems implemented in sub-Saharan Africa and to help educators to find ways of maximising systems usage.