The algal multiphase transient chlorophyll fluorescence kinetic curve (also known as the OJIP curve) is employed to extract differences in distance, shape, and angle features, and curve differences are thereby measured. The K-nearest neighbor (KNN) classification algorithm is used to evaluate those features and select the combination of features with a better classification effect. Weights are assigned according to the contribution rates of different features to toxicity intensity characterization for feature fusion, which is followed by the construction of the characterization parameter PI_FCD of the comprehensive toxicity of water. Moreover, Chlorella vulgaris is adopted as the test algae species to compare PI_FCD with the current characterization parameters Fv/Fm and PI_ABS under the stress of toxic substances (DCMU, DBMIB, MV, Malathion, and Carbofuran). The effectiveness and superiority of the proposed parameter PI_FCD are verified from the aspects of toxicity response, minimum detection limit, and maximum response concentration. The results show that PI_FCD is sensitive to all the five toxic substances and can detect more kinds of pollutants than Fv/Fm. The minimum detection limits of PI_FCD for the five toxic substances are respectively 90.34%, 41.66%, 81.91%, 95.43%, and 77.66% lower than those of PI_ABS. PI_FCD also demonstrates a higher detection ability for high-concentration toxic substances.