如何从破折号下拉菜单中选择和运行模型并更新混淆矩阵图？

Question

我正在构建一个基于这个乳腺癌数据集的ML预测破折号应用程序。

从下拉菜单，我希望能够选择我的模型之一，运行拟合，并返回更新的混乱矩阵(热图)。

我计划将脚本扩展到表、roc曲线、学习-曲线等(即多输出回调 )--但首先，我希望在实现其他元素之前，这个部分能够工作。

我尝试过不同的东西。

例如，在当前代码(如下所示)之前，我尝试从下拉菜单中直接调用模型，然后在回调中执行所有cm计算，得到一个AttributeError：'str‘对象没有属性'fit'

@app.callback(Output('conf_matrix', 'figure'), [Input('dropdown-5', 'value')])
def update_cm_matix(model):
    class_names=[0,1]
    fitModel = model.fit(X_train, y_train)
    y_pred = fitModel.predict(X_test)
    cm = confusion_matrix(y_test, y_pred)
    return {'data': [go.Heatmap(x=class_names, y=class_names, z=cm, showscale=True, colorscale='blues')],
            'layout': dict(width=350, height=280, margin={'t': 10},
                       xaxis=dict(title='Predicted class', tickvals=[0, 1]),
                       yaxis=dict(title='True class', tickvals=[0, 1], autorange='reversed'))}

(替换下面脚本中的app.callback和函数)。

我正在努力的当前版本是：

# -*- coding: utf-8 -*-
import dash
import dash_core_components as dcc
import dash_html_components as html
import dash_bootstrap_components as dbc
import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import GridSearchCV, train_test_split
from sklearn.metrics import confusion_matrix
from sklearn.feature_selection import RFE
import plotly.graph_objs as go
from dash.dependencies import Input, Output

app = dash.Dash(__name__, external_stylesheets=[dbc.themes.BOOTSTRAP])
server = app.server

app.config.suppress_callback_exceptions = True

df = pd.read_csv("breast_cancer.csv")
y = np.array(df.diagnosis.tolist())
data = df.drop('diagnosis', 1)
X = np.array(data.values)

scaler = StandardScaler()
X = scaler.fit_transform(X)

random_state = 42
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=random_state)

# First model: logistic model + optimize hyperparameters
log = LogisticRegression(random_state=random_state)
param_grid = {'penalty': ['l2', 'l1'], 'C': [0.001, 0.01, 0.1, 1, 10, 100, 1000]}
CV_log = GridSearchCV(estimator=log, param_grid=param_grid,, scoring='accuracy', verbose=1, n_jobs=-1)
CV_log.fit(X_train, y_train)
log_best_params = CV_log.best_params_
log_clf = LogisticRegression(C=log_best_params['C'], penalty=log_best_params['penalty'], random_state=random_state)

# Second model: logistic model with recursive features elimination (just for illustration purposes, other models will be included)
rfe_selector = RFE(log_clf)

# app layout
app.layout = html.Div([
    html.Div([
        dcc.Dropdown(
            id='dropdown-5',
            options=[{'label': 'Logistic', 'value': 'log_clf'},
                     {'label': 'RFE', 'value': 'rfe_selector'}],
            value='log_clf',
            style={'width': '150px', 'height': '35px', 'fontSize': '10pt'}
        )], style={}),

    html.Div([
        dcc.Graph(id='conf_matrix')
    ])
])

# function to run selected model
def ClassTrainEval(model):
    fitModel = model.fit(X_train, y_train)
    y_pred = fitModel.predict(X_test)
    cm = confusion_matrix(y_test, y_pred)
    return fitModel, y_pred, y_score, cm

models = [log_clf, rfe_selector]
class_names = [0,1]

# dash callback
@app.callback(Output('conf_matrix', 'figure'), [Input('dropdown-5', 'value')])
def update_cm_matix(model):
    for model in models:
        ClassTrainEval(model)
    return {'data': [go.Heatmap(x=class_names, y=class_names, z=cm, showscale=True, colorscale='blues')],
            'layout': dict(width=350, height=280, margin={'t': 10},
                           xaxis=dict(title='Predicted class', tickvals=[0, 1]),
                           yaxis=dict(title='True class', tickvals=[0, 1], autorange='reversed'))}

if __name__ == '__main__':
    app.run_server(debug=True)

其中我得到了一个NameError:名称'cm‘不是定义的错误。

我真的不知道怎样才能把这件事办好--所以我希望有人能给我指明正确的方向。

谢谢!

Answer 1

代码中存在多个错误。让我们先算出你的两次尝试。

dcc.Dropdown(
        id='dropdown-5',
        options=[{'label': 'Logistic', 'value': 'log_clf'},
                 {'label': 'RFE', 'value': 'rfe_selector'}],
        value='log_clf',
        style={'width': '150px', 'height': '35px', 'fontSize': '10pt'}
    )], style={})

在下拉列表中，模型是一个字符串(type('log_clf') == str)，所以您不能训练它。您需要按以下方式编写回调：

models = {'Logistic':log_clf, 'RFE':rfe_selector}
""""i jumped some line of code"""
dcc.Dropdown(
        id='dropdown-5',
        options=[{'label': v, 'value': v} for v in ['Logistic','RFE']],
        value='Logistic',
        style={'width': '150px', 'height': '35px', 'fontSize': '10pt'}
    )

对于第二次尝试，您还需要一行代码来适应我所做的更改：

错误是：NameError: name 'cm' is not defined error (我假设它发生在回调中)，并且正在发生，因为您没有将函数的输出分配给变量：

其功能是

# function to run selected model
def ClassTrainEval(model):
    fitModel = model.fit(X_train, y_train)
    y_pred = fitModel.predict(X_test)
    cm = confusion_matrix(y_test, y_pred)
    return fitModel, y_pred, y_score, cm #Note that y_score is never defined so you need to remove this 

然后在回调中：

# dash callback
@app.callback(Output('conf_matrix', 'figure'), [Input('dropdown-5', 'value')])
def update_cm_matix(model):
    for model in models: #<-------No loop needed
        ClassTrainEval(model) #<-------Here You need to assigne the output
    return {'data': [go.Heatmap(x=class_names, y=class_names, z=cm, showscale=True, colorscale='blues')],
            'layout': dict(width=350, height=280, margin={'t': 10},
                           xaxis=dict(title='Predicted class', tickvals=[0, 1]),
                           yaxis=dict(title='True class', tickvals=[0, 1], autorange='reversed'))}

你可能想写：

@app.callback(Output('conf_matrix', 'figure'), [Input('dropdown-5', 'value')])
def update_cm_matix(v):
    model = models[v]
    fitModel, y_pred, cm =  ClassTrainEval(model)
    return {'data': [go.Heatmap(x=class_names, y=class_names, z=cm, showscale=True, colorscale='blues')],
            'layout': dict(width=350, height=280, margin={'t': 10},
                           xaxis=dict(title='Predicted class', tickvals=[0, 1]),
                           yaxis=dict(title='True class', tickvals=[0, 1], autorange='reversed'))}