X.reshape() 的问题

Question

I am trying to use predict (as can be seen below) to predict data on a model after training it. However all the steps I take I get an error.

I tried:

X = data2[["SEX", "NAGE", "LIFEAUD5", "PRS"]]
X = X.values.reshape(X.shape)
X = X.transpose()
y_pred=model.predict(X)
n_scores = cross_val_score(model, X,y, scoring='accuracy', cv=cv, n_jobs=-1)
# report the model performance
print('Mean Accuracy: %.3f (%.3f)' % (mean(n_scores), std(n_scores)))

but then get the error: ValueError: X has 17016 features, but LogisticRegression is expecting 4 features as input.

So then I put in the dimensions of the data:

X = X.values.reshape(X.shape(4,17016))
X = X.transpose()
y_pred=model.predict(X)
n_scores = cross_val_score(model, X,y, scoring='accuracy', cv=cv, n_jobs=-1)
# report the model performance
print('Mean Accuracy: %.3f (%.3f)' % (mean(n_scores), std(n_scores)))

Only to be told: TypeError: 'tuple' object is not callable

Then I tried:

X = data2[["SEX", "NAGE", "LIFEAUD5", "PRS"]]
X = X.values.reshape(X.shape[4,17016])
X = X.transpose()
y_pred=model.predict(X)
n_scores = cross_val_score(model, X,y, scoring='accuracy', cv=cv, n_jobs=-1)
# report the model performance
print('Mean Accuracy: %.3f (%.3f)' % (mean(n_scores), std(n_scores)))

But I am still getting an error: tuple indices must be integers or slices, not tuple

So then I make it slices:

X = data2[["SEX", "NAGE", "LIFEAUD5", "PRS"]]
X = X.values.reshape(X.shape[4:17016])
X = X.transpose()
y_pred=model.predict(X)
n_scores = cross_val_score(model, X,y, scoring='accuracy', cv=cv, n_jobs=-1)
# report the model performance
print('Mean Accuracy: %.3f (%.3f)' % (mean(n_scores), std(n_scores)))

but still get an error of: ValueError: cannot reshape array of size 68064 into shape () Which leads me down the same rabbit hole.

I feel like I have gone in circles with all the posts I have been looking at, none which seem to solve my problem.

Here is the head of the data: {'IID': {0: 'NESARC00001', 1: 'NESARC00003', 2: 'NESARC00005', 3: 'NESARC00006', 4: 'NESARC00007'}, 'SEX': {0: 1, 1: 2, 2: 1, 3: 1, 4: 2}, 'NAGE': {0: 79, 1: 49, 2: 76, 3: 51, 4: 36}, 'LIFEAUD5': {0: 0, 1: 0, 2: 0, 3: 0, 4: 1}, 'YES': {0: 'Yes', 1: 'Yes', 2: 'Yes', 3: 'Yes', 4: 'Yes'}, 'PRS': {0: 0.000738316709, 1: -0.0027634842900000002, 2: 0.000455826588, 3: -0.000639326695, 4: -0.0028217993}}

THe dimensions are 17016,6.

update, I have changed the dimensions to 6 11344 but still seem to be getting the same errors.

Answer 1

When calling predict and fit, an array or arraylike object of dimension n_samples x n_features is expected. Looking at those lines

X = data2[["SEX", "NAGE", "LIFEAUD5", "PRS"]]
X = X.values.reshape(X.shape)
X = X.transpose()

You are first extracting an array of n_samples x 4 with n_samples probably being 17016. The second line is not needed and the transpose() in the end is the problem as you are swapping the dimensions. This should explain the error you get. I would assume that y_pred=model.predict(data2[["SEX", "NAGE", "LIFEAUD5", "PRS"]]) should just work as sklearn classes accept both numpy arrays and pandas DataFrames.

Apart from this, X = X.values.reshape(X.shape(4,17016)) gives you an error as X.shape returns a tuple and you are trying to call a method with 2 arguments so to say using braces. This X.shape[4,17016] gives you an error as you are trying to index into it with a wrong syntax. If you want to call reshape, just pass in a tuple of the required and allowed dimensions like X.values.reshape((4,17016)),