This study introduces a weighted k-nearest neighbour model for user positioning in a visible light communications (VLC) system. The new model offers a higher degree of accuracy compared with conventionally existing techniques in VLC, such as trilateration. In the proposed model, the current position of the receiver is estimated based on the positions of the k-NN (known as reference points) pre-defined and recorded in a lookup table. The Euclidean distances from the actual receiver to the reference points are weighted in order to improve accuracy. Simulation results show that the proposed model outperforms the trilateration method by 36 and 50% accuracy with and without ambient light interference, respectively.