教你用PyQt一步步打造一个属于自己的机器学习GUI-初步训练及评估-06

在界面设计完成之后，就需要关联各个控件的事件。当然，完成如下所示的模型评估GUI界面，看似简单，却仍需要做大量重复工作，比如各个分类回归算法参数设计，分类回归算法评估方式及绘图形式。在这里做一些规定，针对结构化数据集，我们希望，数据格式如下：第一列为样本标识，中间为特征列，最后一列为预测目标列。通过单击打开训练数据按钮，展示数据集，同时相关特征展示在描述子选择树形控件下，可供选择。详细的操作见教你用PyQt一步步打造一个属于自己的机器学习GUI-数据加载展示-03。

 

 

import csv
from sklearn.preprocessing import label_binarize
from sklearn.multiclass import OneVsRestClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score, roc_curve, auc, \
    precision_recall_curve
from sklearn.model_selection import cross_val_score, train_test_split, LeaveOneOut, StratifiedKFold
from sklearn.preprocessing import StandardScaler
from sklearn.svm import SVC
import pandas as pd
import numpy as np
from sklearn.metrics import *

class ModelEvaluation(object):

    def __init__(self,X,y):
        self.X = X
        self.y = y

    def holdout_binary_task(self,model,test_size = 0.3):
        metrics_dict = pd.DataFrame()
        X_train, X_test, y_train, y_test = train_test_split(self.X, self.y, test_size = test_size, random_state = 42)
        model.fit(X_train,y_train)
        y_pred = model.predict(X_test)
        y_prob = model.predict_proba(X_test)[:,1]
        accuracy = accuracy_score(y_test, y_pred)
        precision_ = precision_score(y_test, y_pred)
        recall_ = recall_score(y_test, y_pred)
        f1 = f1_score(y_test, y_pred)

        fpr, tpr, _ = roc_curve(y_test, y_prob)
        roc_auc = auc(fpr, tpr)

        precision, recall, _ = precision_recall_curve(y_test, y_prob)
        pr_auc = auc(recall, precision)

        metrics_dict = metrics_dict.append({
            'accuracy': accuracy,
            'precision_': precision_,
            'recall_': recall_,
            'F1-score': f1,
            'AUC-ROC': roc_auc,
            "PR-AUC":pr_auc,
            "fpr":fpr,
            "tpr": tpr,
            "precision": precision,
            "recall": recall,
        },ignore_index=True)

        plot_roc_data = pd.DataFrame({
            "fpr": fpr,
            "tpr": tpr,
        }).to_csv("D://pycharm_project//Machine_Learning_GUI//MachineLearning_GUI_Design//datas//holdout_roc_%s.csv" % (
            model.__class__.__name__))

        plot_pr_data = pd.DataFrame({
            "precision": precision,
            "recall": recall
        }).to_csv("D://pycharm_project//Machine_Learning_GUI//MachineLearning_GUI_Design//datas//holdout_pr_%s.csv" % (
            model.__class__.__name__))

        metrics_dict.to_csv("D://pycharm_project//Machine_Learning_GUI//MachineLearning_GUI_Design//datas//holdout_%s.csv"%(model.__class__.__name__))
        return metrics_dict

    def loocv_binary_task(self,model):
        X_train = np.array(self.X)
        y_train = np.array(self.y)

        loo = LeaveOneOut()
        all_y = []
        all_probs = []
        accuracy = 0
        y_ps = []
        for train, test in loo.split(X_train, y_train):
            model.fit(X_train[train], y_train[train])
            y_p = model.predict(X_train[test])
            y_ps.append(y_p)
            if y_p == y_train[test]:
                accuracy += 1
            all_y.append(np.array(y_train)[test])
            all_probs.append(model.fit(X_train[train], y_train[train]).predict_proba(X_train[test])[:, 1])
        fpr, tpr, thresholds = roc_curve(all_y, all_probs)
        roc_auc = auc(fpr, tpr)

        precision, recall, _ = precision_recall_curve(all_y, all_probs)
        pr_auc = auc(recall, precision)

        precision_ = precision_score(all_y, y_ps)
        recall_ = recall_score(all_y, y_ps)
        f1 = f1_score(all_y, y_ps)

        plot_roc_data = pd.DataFrame({
            "fpr":fpr,
            "tpr":tpr,
        }).to_csv("D://pycharm_project//Machine_Learning_GUI//MachineLearning_GUI_Design//datas//loocv_roc_%s.csv"%(model.__class__.__name__))

        plot_pr_data = pd.DataFrame({
            "precision": precision,
            "recall": recall
        }).to_csv("D://pycharm_project//Machine_Learning_GUI//MachineLearning_GUI_Design//datas//loocv_pr_%s.csv"%(model.__class__.__name__))
        return roc_auc,pr_auc, accuracy/np.shape(X_train)[0],precision_,recall_,f1,fpr,tpr, precision, recall

    def skfcv_binary_task(self,model,k_fold = 5):
        pass

    def holdout_multi_task(self,model,test_size = 0.2,classes_list = [0,1,2]):
       pass

以上代码仅供参考，尚未成熟，同时只是整个GUI界面设计的其中一个环节，后面会整合自动特征提取，特征选择，超参数自动选优，以及材料领域一些深度学习应用如图卷积，电池soc,soh等时序数据预测。欢迎提供建议和意见。

更多内容，参考订阅号：数道