#!/usr/bin/env python # -*- coding: utf-8 -*- import pickle from sklearn.metrics import accuracy_score, precision_score, recall_score from sklearn.metrics import confusion_matrix, classification_report from sklearn.utils import shuffle from sklearn import tree from sklearn import svm from sklearn.neighbors import KNeighborsClassifier from sklearn import cross_validation #import matplotlib.pyplot as plt from sklearn.tree import DecisionTreeClassifier from sklearn.naive_bayes import GaussianNB from sklearn.dummy import DummyClassifier from sklearn import linear_model import numpy as np from sklearn.cross_validation import train_test_split [n,PearsonNumpy, PearsonScipy, Kendall, Spearman, PearsonNumpy, DiccPearsonScipy, DiccKendallScipy, DiccSpearmanScipy, DataSismos, key2vector]= pickle.load( open( "correlaciones_jazminHernan-jazminMeses-volumen.p", "rb" ) ) Grupos = { 'sacudir_orig-magnitud_orig-velocidadpromedio' : ["sacudir_orig","magnitud_orig","velocidadpromedio"], 'sacudir_orig-magnitud_orig' : ["terremoto","chile"], 'temblor_orig-sacudir_orig-magnitud_orig-velocidadpromedio' : ["temblor_orig","sacudir_orig","magnitud_orig","velocidadpromedio"], 'temblor_orig-sacudir_orig-magnitud_orig': ["temblor_orig","sacudir_orig","magnitud_orig"], "sacudir_orig":["sacudir_orig"], "sacudir_orig-velocidadpromedio":["sacudir_orig","velocidadpromedio"], "terremoto_orig-sacudir_orig-magnitud_orig-velocidadpromedio":["terremoto_orig","sacudir_orig","magnitud_orig","velocidadpromedio"], "terremoto_orig-sacudir_orig-magnitud_orig":['terremoto_orig","sacudir_orig","magnitud_orig'], "terremoto_orig-temblor_orig-sacudir_orig-magnitud_orig-velocidadpromedio":["terremoto_orig","temblor_orig","sacudir_orig","magnitud_orig","velocidadpromedio"] } # A Continuación se va a buscar la mejor via de clasificacion, para eso lo que se hace es # ir probando con los grupos de terminos como atributo, primero solo el mas importante, luego los # dos mas importantes y asi R = [] Attr = [] y = [str(x) for x in key2vector['magnitud_real']] y_enteros = [int(float(x)) for x in y] cont = 0 for g in Grupos: R.append([]) Attr.append(g) atts = ",".join(Attr) print '--------------------------------------------------------------------------------' print 'Probando Atributos: ',Attr Z = [] for a in Attr: Z.append(key2vector[a]) X = [list(x) for x in zip(*Z)] for k in range(2,9): #####"knn con k=",k,"clases reales" #knn = KNeighborsClassifier(n_neighbors=k) #knn.fit(X,y) #y_pred = knn.predict(X) #acc = accuracy_score(y,y_pred) #print "knn con k=",k,"clases reales:",acc #R[cont].append(tuple([acc,"knn ("+atts+") con k="+str(k)+"clases reales"])) ##### knn con k=",k,"clases enteras" knn = KNeighborsClassifier(n_neighbors=k) knn.fit(X,y_enteros) y_pred = knn.predict(X) acc = accuracy_score(y_enteros,y_pred) print "knn con k=",k,"clases enteras:",acc R[cont].append(tuple([acc,"knn ("+atts+") con k="+str(k)+"clases enteras",str(k)+"-nn "+atts])) #####"cross_validation de knn con k=",k,"clases reales" #knn = KNeighborsClassifier(n_neighbors=k) #cv_arr = cross_validation.cross_val_score(knn,X,y,cv=10) #valor = sum(cv_arr)/len(cv_arr) #print "knn - cross_validation con k=",k,"clases reales:",valor #R[cont].append(tuple([valor,"knn ("+atts+")- cross_validation con k="+str(k)+"clases reales"])) ##### cross_validation de knn con k=",k,"clases enteras: knn = KNeighborsClassifier(n_neighbors=k) cv_arr = cross_validation.cross_val_score(knn,X,y_enteros,cv=10) valor = sum(cv_arr)/len(cv_arr) print "knn - cross_validation con k=",k,"clases enteras:",valor R[cont].append(tuple([valor,"knn ("+atts+")- cross_validation con k="+str(k)+"clases enteras",str(k)+"-nn "+atts])) ########################################## ### SVM # http://scikit-learn.org/stable/modules/svm.html ## SVC -ovo- real #clf = svm.SVC(decision_function_shape='ovo') #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "svm.SVC(ovo) - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"svm.SVC - ovo -("+atts+")- cross_validation clases reales"])) ## SVC -ovr- real #clf = svm.SVC(decision_function_shape='ovr') #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "svm.SVC(ovr) - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"svm.SVC - ovr -("+atts+")- cross_validation clases reales"])) ## SVC -ovo- entero clf = svm.SVC(decision_function_shape='ovo') scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "svm.SVC(ovo) - cross_validation clases entero:",valor R[cont].append(tuple([valor,"svm.SVC - ovo -("+atts+")- cross_validation clases entero","svm "+atts])) ## SVC -ovr - entero clf = svm.SVC(decision_function_shape='ovr') scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "svm.SVC(ovr) - cross_validation clases entero:",valor R[cont].append(tuple([valor,"svm.SVC - ovr -("+atts+")- cross_validation clases entero","svm "+atts])) ## LinearSVC - real #clf = svm.LinearSVC() #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "LinearSVC - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"svm.LinearSVC ("+atts+")- cross_validation clases reales"])) # LinearSVC -entero clf = svm.LinearSVC() scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "LinearSVC - cross_validation clases enteros:",valor R[cont].append(tuple([valor,"svm.LinearSVC ("+atts+")- cross_validation clases enteros","linearSCV "+atts])) #---- BaseLine contant rales #-- #strategy = "uniform" #clf = DummyClassifier(strategy=strategy) #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "DummyClassifier(uniform) - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"DummyClassifier(uniform) - ("+atts+")- cross_validation clases reales"])) #- #strategy = "most_frequent" #clf = DummyClassifier(strategy=strategy) #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "DummyClassifier(most_frequent) - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"DummyClassifier(most_frequent) - ("+atts+")- cross_validation clases reales"])) #- #strategy = "stratified" #clf = DummyClassifier(strategy=strategy) #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "DummyClassifier(stratified) - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"DummyClassifier(stratified) - ("+atts+")- cross_validation clases reales"])) #-- enteros strategy = "uniform" clf = DummyClassifier(strategy=strategy) scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "DummyClassifier(uniform) - cross_validation clases enteros:",valor R[cont].append(tuple([valor,"DummyClassifier(uniform) - ("+atts+")- cross_validation clases enteros","BaseLine(uniform) "+atts])) #- strategy = "most_frequent" clf = DummyClassifier(strategy=strategy) scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "DummyClassifier(most_frequent) - cross_validation clases enteros:",valor R[cont].append(tuple([valor,"DummyClassifier(most_frequent) - ("+atts+")- cross_validation clases enteros","BaseLine(most_frequent) "+atts])) #- strategy = "stratified" clf = DummyClassifier(strategy=strategy) scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "DummyClassifier(stratified) - cross_validation clases enteros:",valor R[cont].append(tuple([valor,"DummyClassifier(stratified) - ("+atts+")- cross_validation clases enteros","BaseLine(stratified) "+atts])) ############################################## ## Decision Tree # # reales #clf = DecisionTreeClassifier(random_state=0) #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "DecisionTreeClassifier - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"DecisionTreeClassifier - ("+atts+")- cross_validation clases reales"])) # enteros clf = DecisionTreeClassifier(random_state=0) scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "DecisionTreeClassifier - cross_validation clases enteros:",valor R[cont].append(tuple([valor,"DecisionTreeClassifier - ("+atts+")- cross_validation clases enteros","DecisionTreeClassifier "+atts])) ################################################# #### Naive Bayes clf = GaussianNB() #scores = cross_validation.cross_val_score(clf, X, y, cv=10) #valor = sum(scores)/len(scores) #print "GaussianNB - cross_validation clases reales:",valor #R[cont].append(tuple([valor,"GaussianNB - ("+atts+")- cross_validation clases reales"])) # enteros clf = GaussianNB() scores = cross_validation.cross_val_score(clf, X, y_enteros, cv=10) valor = sum(scores)/len(scores) print "GaussianNB - cross_validation clases enteros:",valor R[cont].append(tuple([valor,"GaussianNB - ("+atts+")- cross_validation clases enteros","NaiveBayes "+atts])) ################################################# #### Regresion Lineal regr = linear_model.LinearRegression() y_reales = [float(x) for x in y] X_train, X_test, y_train, y_test = train_test_split(X,y_reales, test_size=0.20) regr.fit(X_train, y_train) media = np.mean((regr.predict(X_test) - y_test) ** 2) print("Mean squared error: %.2f" % media) varianza = regr.score(X_test, y_test) R[cont].append(tuple([media,"Regresion Lineal -Mean:"+str(media)+" varianza:"+str(varianza)+" ("+atts+")"]))#,"NaiveBayes "+atts])) cont+=1