将用户定义的 function 应用于 python dataframe 中特定列的所有行

Question

I have hard times to apply a user defined function to a specific column in a python dataframe. The dataframe is as fellow:

Year    state   Narrative
----------------------------------
2015      WV   a roof fall occurred at 10:05 am at 10+50 entry 6 in 8lms mmu 010, .. more text
2016      AL   a rib rolled out striking him on his left foot resulting ...... more text
2017      CO   a non-injury mountain bump occurred inby the 5n longwall. additional ... more text

I want to predict the type of ground failure based on "Narrative", such that a new column is added to the dataframe as shown below. I predict the ground fall through looking for some keywords in the "narrative", for example: if the "narrative" includes any of the following words ['roof fall', 'roof broke', 'rock fell from the top'] , the ground fall prediction should be "roof fall".

This is the user defined function that I generated, but it is not working.

def predict_groundFall(narrative):
    fall_dict = {'roof fall': ['Roof fall', 'roof broke', 'rock fell from the top'],
                 'rib fall': ['rib fall ', 'rib rolled', 'rib dislodged'],
                 'outburst': ['outburst', 'bounce', 'rockburst']}
    for key, values in fall_dict.iteritems():
        if values in narrative:
            return key
            break
df['predicted_failure'] = df.apply( lambda row:  predict_groundFall( row['Narrative']), axis=1)

this is what I want to achieve: adding a new column to predict the failure from the narrative.

Year    state   Narrative                                        predicted_failure
------------------------------------------------------------- ---------------------
2015      WV   a roof fall occurred ....... more text....                roof fall
2016      AL   a rib rolled out striking ......more text ....             rib fall
2017      CO   a non-injury mountain ....... more text....                 outburst

I am new to Python, so I hope you help me fix the code to make it work. A better method to achieve my goal is highly appreciated. thank you in advance,

Answer 1

Your function isn't working as expected. You want to try the following:

def predict_groundFall(narrative):
    fall_dict = {'roof fall': ['Roof fall', 'roof broke', 'rock fell from the top'],
                 'rib fall': ['rib fall ', 'rib rolled', 'rib dislodged'],
                 'outburst': ['outburst', 'bounce', 'rockburst']}
    for key in fall_dict:
        if any(v.lower() in narrative.lower() for v in fall_dict[key]):
            return key

Then change your column assignment to the following:

df['predicted_failure'] = df["Narrative"].apply(lambda x: predict_groundFall(x))

Answer 2

I think the problem is in your apply function.

change this line df['predicted_failure'] = df.apply( lambda row: predict_groundFall( row['Narrative']), axis=1)

to

df['predicted_failure'] = df.Narrative.apply(predict_groundFall)

this will send each value of Narrative to your custom function and then populate the new column with the return from that function