This is the case where the order of column-names while model building is different from order of column-names while model scoring.
I have used the following steps to overcome this error
First load the pickle file
model = pickle.load(open("saved_model_file", "rb"))
extraxt all the columns with order in which they were used
cols_when_model_builds = model.get_booster().feature_names
reorder the pandas dataframe
pd_dataframe = pd_dataframe[cols_when_model_builds]
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…