Complete the following function that takes in a frame and a list of columns to ignore. For all columns that are not ignored, extract a number

from the cell's text and replace the cell's contents with that correctly typed number $($either a float or int $).$ The number should be

converted to an int if the text does not have a decimal point $($and is not a percentage$).$ If the cell contains no number, replace it with a

numpy. nan $.$ Convert percentages to normalized floats, e$.$g$.,$ str$(\text{'}37\mathrm{.}6\%(2017$ est.$)\text{'})$ turns into float$(0\mathrm{.}376).$ Return the passed$-$

in frame.

For cells with ambiguous data$/$numbers$,$ it is your job to do your best to honor what that data's author likely intended.

Warning: This is a hard task with no exact answer $($as with most tasks in machine learning$).$ Look through the data to find as many

edge cases as you can, turn those into test cases, and see how many you can solve. Start with a simple solution and work up from

there.

def extract$\_$numbers$($frame$,$ ignore$\_$columns $=[])$:

return NotImplemented

print$("$Cleaning up the numeric values:"$)$

extract$\_$numbers$($world$\_$data, $[\text{'}$Country$\text{'},$ 'Export commodities'$])$

As a consequence of what we just did, we also standardized all the NaN values in those columns. Before we could see empty values

represented in several different ways:

$\text{'}$ NaN $\text{'}$

$\text{'}$NA$\text{'}$

'total: NA$\text{'}$

$"($empty

THis is python$3$

Now, since any cell that didn't contain a number was replaced with numpy. nan $,$ all numeric cells with missing$/$empty values are consistent.

Now that we have cleaned up our numeric columns, let's officially tell Pandas the data types for our columns. Because even though most of

our columns now only contains numbers $($and numpy. nan $),$ Pandas still needs to be told what data type each column uses.