<p>The octanol–water partition coefficient (<i>K</i>_OW) is an effective parameter for determining a substance’s chemical and biological activity, as it describes its distribution between polar and non-polar media. The <i>K</i>_OW is often determined by empirical tests but this study predicted it using effective molecular descriptors that describe the molecular structure well. Effective molecular descriptors in determining the log₁₀<i>K</i>_OW of 43 anionic surfactants were predicted by an artificial neural network (ANN) technique and modeled using the quantitative structure–properties relationship (QSPR) framework. First, seven effective descriptors were selected among the 19 calculated descriptors and then an ANN with different numbers of neurons in the hidden layer from 2 to 16 layers was trained using sigmoid transfer function to predict the output values. Based on the statistical results, the network with seven neurons in the input layer, six neurons in the hidden layer, and one neuron in the output layer (structure 7-6-1) was selected as the optimal network with a higher correlation coefficient of 0.924 and lower sum of square error of 0.049. The model determined that all three categories of topological, physical properties, and geometrical descriptors contribute to determining the <i>K</i>_OW value of anionic surfactants. Effective descriptors on the log₁₀<i>K</i>_OW were in order of their influence coefficient melting point (0.162), Van der Waals molecular volume (0.161), triple bond index (0.145), molecular mass (0.145), double bond index (0.143), number of aromatic bonds (0.135), and number of oxygen atoms (0.109). The experimental and predicted log₁₀<i>K</i>_OW values with <i>R</i>² = 0.986 confirmed that the developed ANN can accurately predict the log₁₀<i>K</i>_OW value of new anionic surfactants.</p>