Various Machile Learning Models are trained using scikit-learn on the dataset provided by this paper to predict partition coefficients logP(oct/water).
RDkit was used to extract feature from SMILES string. There are total of 13 features.
There are 6 models in total trained using scikit-learn module: KNN, Ridge, Lasso, Kernelized SVM, Random Forests and MLP. The MLP model with a 1000 neurons layer has the highest score with minimal variance.