The major diagnostic sleep laboratory tool for assessing excessive daytime sleepiness (EDS), the multiple sleep latency test (MSLT), is increasingly criticized for poor precision in the differentiation of idiopathic hypersomnia (IH) and narcolepsy ( Trotti et al., 2013 , Johns, 2000 ). Recent evidence suggests that actigraphy can supplement the diagnostic process by providing information about the sleep-wake rhythm ( Kretzschmar et al., 2016 , Filardi et al., 2015 , Bruck et al., 2005 ). An actigraphy analysis tool is introduced that processes actigraphy recordings with machine learning methods. It is preliminarily validated for recordings of hypersomnolent patients. Sleep is detected using a support vector machine (SVM) with 61 features describing the actigraphy signal. Testing the prediction against polysomnographically validated labels on nocturnal recordings of hypersomnolent patients showed an accuracy of 87%, a sensitivity of 93% and a specificity of 51%. Post-processing the prediction with morphological operations increases the specificity to 73%, with only minor declines in accuracy and sensitivity (85% and 88%, respectively). Individually training an SVM per patient revealed the most promising results with 95% accuracy, 73% specificity and 99% sensitivity. Each detected sleep phase is automatically verified in order to not be confused with episodes of unattached actigraphy device. Moreover, an SVM is trained to identify the main sleep episode of each day, based on features describing a sleep phase’s probability of being a nap. Testing with manually generated labels of 14-day actigraphies of EDS patients showed that 90% of the sleep phases receive the correct label (diurnal or nocturnal sleep), with a specificity of 96% and a sensitivity of 86%. Finally, the classifiers’ annotations are used to calculate 217 sleep-wake rhythm and motor activity parameters. Comparing the automatically derived sleep-wake parameters of 15 long-term actigraphies from IH patients to previously published results ( Filardi et al., 2015 ) showed good accordance. Altogether, the developed tool and the methods used appear a promising pathway for future investigations in the search for disorder inherent sleep-wake rhythm and motor activity patterns in big patient cohorts.