尝试针对我的随机 forrest 模型进行测试时，如何修复“特征数量错误”？

Question

I have a trained model. I want to find out which class a new data belongs to. I've done some trials, but I've encountered some problems.

with open('text_classifier', 'rb') as training_model:
model = pickle.load(training_model)
y_pred2 = model.predict(X_test)

This code works

But...

 new_test_data=["spor toto süper lig 30. hafta medipol bu akşam ev göztepe 
ile saat 20.30'da başla mücadele suat arslanboğa arslanboğa yardımcı 
serka ok ve ismail şencan"]
tfidfconverter = TfidfVectorizer()
new_test_data = tfidfconverter.fit_transform(new_test_data).toarray()
model.predict(new_test_data)

I get an error like this

Number of features of the model must match the input. Model n_features is 9671 and input n_features is 25

The code block I'm training with

data = load_files(r"...\docs",encoding="utf-8")
X, y = data.data, data.target
tfidfconverter = TfidfVectorizer(min_df=3, max_df=0.7)
X = tfidfconverter.fit_transform(X).toarray()
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, 
random_state=0)
classifier = RandomForestClassifier(n_estimators=1000, random_state=0)
classifier.fit(X_train, y_train)
y_pred2 = classifier.predict(X_test)

Answer 1

I believe that you will need to specify the parameters in your data that you want to actually use as features when training the model. It looks like your training model is using the row entries as the features instead of each column. This can be fixed by reading in the data and then turning around and converting it to CSV and then reading it in again. However, this step should be unnecessary if you already know how the data is structured. Basically, you just need to know the names of the columns of the data. You will need the Pandas module for this method. Here is some code...

    data = load_files(r"...\docs",encoding="utf-8")
    data.to_csv('train_data.csv', encoding = 'utf-8', index = False)

then read the training data back in from the CSV...

    train_data = pd.read_csv('train_data.csv')

Now when you call the train_test_split method you should specify what you want to use as the features in the data. This is generally the columns in a data table as these are the metrics being collected to analyze. I define functions to split the data and build the model specifying the features because I think it is easier to understand but you can also just call the functions directly.

    def split_dataset(dataset, train_percentage, feature_headers, target_header):
        train_x, test_x, train_y, test_y = train_test_split(dataset[feature_headers], 
        dataset[target_header], train_size = train_percentage)
        return train_x, test_x, train_y, test_y

    def random_forest_classifier(features, target):
        model = RandomForestClassifier(n_estimators = 500, oob_score = True, n_jobs 
        =-1,random_state = 1, min_impurity_decrease = .01)
        model.fit(features, target)
        return model

Now you are ready to call the functions using your data.

    train_x, test_x, train_y, test_y = split_dataset(train_data, 0.80, 
    train_data.columns[0:24], train_data.columns[-1])

    trained_model = random_forest_classifier(train_x,train_y)

You should now be able to predict against your trained model using the 25 features.