predict_meta.py

from sklearn.preprocessing import MinMaxScaler
import numpy as np
from joblib import dump, load


# Model="knn.joblib"
Model = "meta_models/fr.joblib"
Scaler = "meta_models/scaler.joblib"


Time_Model = "meta_models/adaboost.joblib"

classes_og = [
    'sklearn.KNeighborsClassifier',
    'sklearn.GaussianProcessClassifier',
    'sklearn.DecisionTreeClassifier',
    'sklearn.RandomForestClassifier',
    'sklearn.AdaBoostClassifier',
    'sklearn.GaussianNB',
    'sklearn.QuadraticDiscriminantAnalysis',
    'sklearn.GradientBoostingClassifier',
    'sklearn.LinearDiscriminantAnalysis',
    'sklearn.Perceptron',
    'sklearn.LogisticRegression',
    'sklearn.ComplementNB',
    'sklearn.SVC']

classes = ['K Nearest Neighbors',
           'Gaussian Process',
           'Decision Tree', 'Random Forest',
           'AdaBoost', 'Gaussian NB',
           'QDA',
           'Gradient Boosting',
           'LDA', 'Perceptron',
           'Logistic Regression', 'Complement NB',
           'SVC']


excluded = [
    'Gaussian Process',
    'Perceptron',
    'Logistic Regression',
    'Complement NB']


def filter_excluded(ls):
    res = []
    for ent in ls:
        if ent[0] in excluded:
            res.append([ent[0], float(0)])
        else:
            res.append(ent)
    return res


def predict_meta(meta):
    model = load(Model)
    scaler = load(Scaler)
    #meta = meta[~np.isnan(meta)] 
    X = scaler.transform(meta.reshape(1, -1))
    outp = model.predict_proba(X)[0]
    srt = np.argsort(outp)[::-1]
    ress = [[classes[srt[i]], float(outp[srt[i]])] for i in range(len(srt))]
    ress = filter_excluded(ress)  # comment this out to get full results
    return ress


def predict_time(meta):

    # model=load(Time_Model)
    # X=meta[[0,2]].reshape(1,-1)
    # outp=model.predict(X)
    # print(meta)
    print(meta[0], meta[2])
    outp = meta[0] * meta[2] / 200
    return outp