Fig. 5
From: OWSum: algorithmic odor prediction and insight into structure-odor relationships
Overview of the classification algorithm OWSum. A Input matrix: Objects Ok can have different features Fj and belong to one or more classes Ci. If a feature occurs to an object or if the object belongs to a class, the values of the cells are 1 otherwise 0. B Calculation of the influence values using Eqs. 1 and 2 and the tf-idf value (Eq. 3) as weighting factor (compare step 2) C Calculation of the score the prediction is based on (Eq. 4, compare step 3)