python svm

 

def execute(ifn, ofn, label, tunner):

    ofh = open(ofn, "w+")

    list_data, list_target = load_data(ifn, tunner)

    ofh.write("all:%d\n" % len(list_data))

 

    list_c=[0.5, 1, 2, 5]

 

    for float_c in list_c:

        ofh.write(".......%f\n" % float_c)

        clf = svm.SVC(kernel='linear', C=float_c)

        data_train = pandas.DataFrame(list_data, columns=list_key)

        sores = cross_validation.cross_val_score(clf, data_train[list_train], list_target, cv=3)

        ofh.write("%s\n" % ",".join(map(str, sores)))

 

        list_train_data, list_test_data, list_train_target, list_test_target = train_test_split(list_data, list_target, test_size=0.3)

        ofh.write("train:%d, test:%d\n" % (len(list_train_data), len(list_test_target)))

        data_train = pandas.DataFrame(list_train_data, columns=list_key)

        data_test = pandas.DataFrame(list_test_data, columns=list_key)

        clf.fit(data_train[list_train], list_train_target)

    #    joblib.dump(clf, 'filename.pkl%s.%s' % (str(float_c), label)) 

 

        res = clf.predict(data_test[list_train])

        total = len(res)

        right = 0

        for l,r in zip(res, list_test_target):

            if l == r:

                right += 1

        ofh.write("total:%f right:%f radio:%f\n" % (total, right, float(right) * 100 /total))

        ofh.write("[\n")

        list_coef = clf.coef_.tolist()

        ofh.write("%s\n" % "\n".join(map(str, [-1 * item for item in list_coef[0]])))

        ofh.write("]\n")

        ofh.write("...................\n")

    ofh.close()