The invention provides a human behavior dependency degree analysis method based on a neural network, which comprises the following steps: S1, processing individual behavior data according to selectedgranularity to obtain daily behavior vectors of tested individuals; S2, according to the current date and individual specific behavior habits, obtaining daily individual vectors of all tested individuals; S3, taking the behavior vector of the tested individual as the public structure input of the neural network, taking the individual daily individual personalized vector as the private structure input of the neural network, taking the daily evaluated behavior time of the individual as the target output, setting the number of hidden layer units, the learning rate and the training round number, and initializing the neural network parameters; S4, training the neural network, and updating neural network parameters by using a standard back propagation algorithm; and S5, the evaluated behavior characteristics of each individual being corresponding to at least one neural unit, and taking the sum of contributions of the evaluated behavior characteristics to all hidden layer units as a specificbehavior dependence degree value of the individual, thereby forming an objective quantitative analysis result.