ValueError: Unable to coerce to Series, length must be 1: given n

Question

I have been trying to use RF regression from scikit-learn, but I’m getting an error with my standard (from docs and tutorials) model. Here is the code:

import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor

db = pd.read_excel('/home/artyom/myprojects//valuevo/field2019/report/segs_inventar_dataframe/excel_var/invcents.xlsx')

age = df[['AGE_1', 'AGE_2', 'AGE_3', 'AGE_4', 'AGE_5']]

hight = df [['HIGHT_','HIGHT_1', 'HIGHT_2', 'HIGHT_3', 'HIGHT_4', 'HIGHT_5']]

diam = df[['DIAM_', 'DIAM_1', 'DIAM_2', 'DIAM_3', 'DIAM_4', 'DIAM_5']]

za = df[['ZAPSYR_', 'ZAPSYR_1', 'ZAPSYR_2', 'ZAPSYR_3', 'ZAPSYR_4', 'ZAPSYR_5']]

tova = df[['TOVARN_', 'TOVARN_1', 'TOVARN_2', 'TOVARN_3', 'TOVARN_4', 'TOVARN_5']]

#df['average'] = df.mean(numeric_only=True, axis=1)


df['meanage'] = age.mean(numeric_only=True, axis=1)
df['meanhight'] = hight.mean(numeric_only=True, axis=1)
df['mediandiam'] = diam.mean(numeric_only=True, axis=1)
df['medianza'] = za.mean(numeric_only=True, axis=1)
df['mediantova'] = tova.mean(numeric_only=True, axis=1)

unite = df[['gapA_segA','gapP_segP', 'A_median', 'p_median', 'circ_media','fdi_median', 'pfd_median', 'p_a_median', 'gsci_media','meanhight']].dropna()

from sklearn.model_selection import train_test_split as ttsplit

df_copy = unite.copy()
trainXset = df_copy[['gapA_segA','gapP_segP', 'A_median', 'p_median', 'circ_media','fdi_median', 'pfd_median', 'p_a_median', 'gsci_media']]

trainYset = df_copy [['meanhight']]

trainXset_train, trainXset_test, trainYset_train, trainYset_test = ttsplit(trainXset, trainYset, test_size=0.3) # 70% training and 30% test

rf = RandomForestRegressor(n_estimators = 100, random_state = 40)
rf.fit(trainXset_train, trainYset_train)
predictions = rf.predict(trainXset_test)
errors = abs(predictions - trainYset_test)
mape = 100 * (errors / trainYset_test)
accuracy = 100 - np.mean(mape)
print('Accuracy:', round(accuracy, 2), '%.')

But output doesn’t look ok:

---> 24 errors = abs(predictions - trainYset_test)
     25 # Calculate mean absolute percentage error (MAPE)
     26 mape = 100 * (errors / trainYset_test)
..... somemore track
ValueError: Unable to coerce to Series, length must be 1: given 780

How can I fix it? 780 is the shape of trainYset_test. I’m not asking for a solution (i.e. write code for me), but for advice on why this error happened. I followed everything as in tutorials.

Answer 1

by seeing in error it is cleared that, the array has to have the shape of one ,

so use reshape to make it in correct shape,

predictions=predictions.reshape(780,1)

Answer 2

I solved this by making sure the predictions were the same data type as the actual data. In my case, it was:

MSE = (sum((y_test-predictions)**2))/(len(newX)-len(newX.columns))

I resolved this by casting y_test to be a numpy array:

MSE = (sum((np.array(y_test)-predictions)**2))/(len(newX)-len(newX.columns))