Data complexity has increased manifold in the age of data-driven societies. The data has become huge and inherently complex. The single-label classification algorithms that were discrete in their operation are losing prominence since the nature of data is not monolithic anymore. There are now cases in machine learningwhere data may belong to more than one class or multiple classes. This nature of data has created the need for new algorithms or methods that are multi-label in nature. Label distribution learning (LDL) is a new way to view multi-labelled algorithms. It tries to quantify the degree to which a label defines an instance. Therefore, for every instance there is a label distribution. In this paper, we introduce a new learning method, namely, angular label distribution learning (ALDL). It is based on the angular distribution function, which is derived from thecomputation of the length of the arc connecting two points in a circle. Comparative performance evaluation in terms of mean-square error (MSE) of the proposed ALDL has been made with algorithm adaptation of k-NN (AA-kNN), multilayer perceptron, Levenberg–Marquardt neural network and layer-recurrent neural network LDL datasets. MSE is observed to decrease for the proposed ALDL. ALDL is also highly statistically significant for the real world datasets when compared with the standard algorithms for LDL.