- Why Pandas?
- Installing Pandas
- Getting the data used in this video
- Loading the data into Pandas (CSVs, Excel, TXTs, etc.)
#####--- adding (10 not included) columns to new column Total ---####li_columns = df.columns.tolist()df['Total'] = 0
for element in li_columns[]: #From HP to Speeddf['Total'] = df['Total'] + df[element]print(df.head(5)) #Print first 5 rows
Just wanna make sure that this one doesn’t go ignored haha @
- Reading Data (Getting Rows, Columns, Cells, Headers, etc.)
, why the double third bracket in [[ 'Name' , 'Type 1' , 'HP']] ? Why not just [ 'Name' , 'Type 1' , 'HP'] ?
- Iterate through each Row
For , refer https://stackoverflow.com/questions/16476924/how-to-iterate-over-rows-in-a-dataframe-in-pandas
usefor index, row in df.iterrows():print(index, row['Name'])
- Getting rows based on a specific condition
In , what if i want to select both 'Grass' and 'Fire', could you show me how can I do that?
- High Level description of your data (min, max, mean, std dev, etc.)
- my bookmark so I can return with the stats function
- Sorting Values (Alphabetically, Numerically)
- Making Changes to the DataFrame
- Adding a column
If You have too many columns to add, as he has at and even more, use this:
- Deleting a column
at you could do like this :df.drop(columns=['Total'], inplace=True)
- Summing Multiple Columns to Create new Column.
Hi Keith,Amazing work. thanks for the Tutorial,however I got an error after i execute the .iloc method at in the video.Can you pls explain how to get rid of the error "Indexing error:Too many Indexers".Thanks in advance
at when i put pokemon['Total'] = df.iloc[:, 4:10].sum(axis=1), bulbasaur's total is 591. but when i put pokemon['Total'] = df.iloc[:, 4:5].sum(axis=1), the total is correct at 318. What happened here?
- Rearranging columns
Best Pandas tutorial on YouTube, especially
XD *SPAM IT*
to 24:33 is essential. The other stuff is good too.
Hi Keith, this is a very helpful introductory course for pandas and thanks a lot! I see a very interesting edit to your video though - at there is `cols = list(df.columns.values)` but a second later at 26:40 it became `cols = list(df.columns)`. What happened here? Why did you remove the `.values`?
The correct or a better approach for mistake in . Will be to use cols[-1:]
Hey, thank You very much for this tutorial, but I'm wondering what happens with "Total" column on ? Numbers became more than twice bigger.
why the total changes? the first row was 318, but in this moment change to 864
It seems that the dataframe got scrambled up a bit there, most likely from having the cell running multiple times. Even when there was an error message, it appears that either the Total or the Legendary column was moved to the left of HP. Upon running the cell again (with the corrected version?) it calculated a new Total adding the previous values and generating corrupted results.
I really appreciate your work, thank you so much :) - bookmark
Very nice tutorial. Might be worth it to demo how to reorder columns by using indices as you allude to starting around . That would be something a newbie might like to see illustrated
- Saving our Data (CSV, Excel, TXT, etc.)
the totals are all messed up and way too high though
there was a bug, each time you run your code the columns will be in different indexes and that because you set the rearrange line to the same variable which is "df", I solved this problem by setting a new variable to the rearrange line then save it to that new variable, each time I run the code now the result is the same.
is where he's getting texts from actual pandas
@Kevin
- Filtering Data (based on multiple conditions)
You don't have to use 'loc' for filtering
When I try to do the operation at I get a warning "future warning:element wise comparison failed, returning scalar instead, but in the future will perform element wise comparison"
Hi Keith! thanks for this helpful video. How do you do the filtering data with different conditions (i.e. at min in the video) for a data frame without headers and just based on position. I tried your example with iloc and it doesn't work. I searched for the solution but was not successful. I hope you see my comment and answer it pls! Thanks in advance for your help.
I think Pandas and NumPy (When using arrays) implement a "C" language style syntax when doing boolean and a few data structures.
I tried to do it with and operator but I got an Error.I tried it with & operator but I got a KeyError: FalseI did it in the PyCharm. Therefore, should I install Anaconda and use it instead of PyCharm?
- Reset Index
At the "Reset index" part in , even after applying two methods -1) new_df = new_df.reset_index(drop=True)new_df2)new_df.reset_index(drop=True,inplace=True)new_dfMy old index column was still in the data set.So i just normally deleted that column by '
- Regex Filtering (filter based on textual patterns)
i was looking for .str.contains('*') filterthanks
I have one comment though. When you run (at )
At flags = re.I parameter in str.contains function didn't work. Looked it up, and it should read flags = re.IGNORECASE. Possibly this is due to a version change?
just one correction: at 'pi[a-z]*' the star means zero or more, instead of 'one or more' as said by Keith.
At , the following may be simpler: new_df = df.loc[df['Name'].str.contains('^pi', regex=True, flags=re.I)]
- Conditional Changes
' after 'Fire' at ?
When i do multiple filtering get this error: SettingWithCopyWarning: A value is trying to be set on a copy of a slice from a DataFrame. Try using .loc[row_indexer,col_indexer] = value instead. Any suggestion ?
- Aggregate Statistics using Groupby (Sum, Mean, Counting)
at what happens with the "Type 1" column label? I tried it here with my own data set and it seems to get lost as a label that I can query for.. How do I group exactly like this but keep the column label as is?
At , how can I plot the data grouped by types , where the type with the highest total is shown from the left to right? Without losing the mapping between the type and its corresponding total value? In my barplot graph, the order doesn´t make sense when I add: .sort_values("Total", ascending = False).
You can also just write the code: df.groupby(['Type 1']).count()['Name']
you can use .size() to get the count of each Pokemon type instead of adding a new column.It would look like this:df.groupby(['Type 1']).size()
