Several researchers have paid attention on locating coordinates of sensors due to its importance for many applications in Wireless Sensor Networks (WSNs). Usually WSNs consist of a number of small, limited energy and low processing capabilities sensors. These sensors can communicate with each other to carry out a specific task according to the application they have been designed for. Using GPS embedded with each sensor may be expensive in terms of cost and consumed energy. So the attention is focused toward designing localization algorithms without using more GPSs. In this paper, machine-learning approach has been proposed to estimate the position of each unknown node in WSNs. It is based on signal strength and positions of three anchors (sensors know their positions). K-Nearest Neighbours KNN and Neural Network (perceptron back propagation) are used in the proposed model to estimate the coordinates of unknown sensors based on data collected from simulation. The main idea of this paper is motivated by a hypothesis which assumes that the position of sensor has a strong correlation with the signal intensity of other nearby sensors. The results emphasise a good accuracy obtained using the model comparing with some related works. Indeed, the average of difference in distances between the actual locations and the estimated locations of the sensor nodes is about 0.84 m.