Preprocessing data for Machine Learning - Python Programming for Finance p. 9

Hello and welcome to part 9 of the Python for Finance tutorial series. In the previous tutorials, we've covered how to pull in stock pricing data for a large number of companies, how to combine that data into one large dataset, and how to visually represent at least one relationship between all of the companies. Now, we're going to try to take this data and do some machine learning with it!

[fuba]

I cannot express how much I enjoy your videos. Thank you for making them!

[hankblack]

Just discovered your Python Finance series two days ago and am working along on a different screen while watching.
Thank you so much - the videos are very informative.

[dmitrypetrov]

thank you for the videos! However, I would suggest to take log returns as they are more linear than classical returns.

[alie]

hi, thanks for everything.
one point;what u calculated is not percentage.

[hajaksksnsjksksbsnsn]

6:29 I think divided by should be previous day price, not current price.

[adeshmallaHQ]

I feel like there is bias in the data [Data Snooping or Look ahead].
The percentage change for a one day look back period shows up earlier than we might have access to the data and the same continues for others.

for example
df = pd.read_csv('sp500_joined_closes.csv', parse_dates=True, index_col=0)
df['AAPL_rets'] = df['AAPL'].pct_change() df['AAPL_1d'] = (df['AAPL'].shift(-1) - df['AAPL'])/df[AAPL'] should have the same result.

I'm Confused each look back period assumes that the future data is already available for use which is a huge bias. Please correct me if I'm wrong

[tomvkgames]

I still don"t understand the shift part in line: df["{}_{}d".format(ticker, i)] = (df[ticker].shift(-i) - df[ticker])/ df[ticker]

First question: You said that: df["{}_{}d".format(ticker, i)]
is the Adj Close value for i days in the future. But you also don't know df[ticker].shift(-i), because that just gives NaN because future data is not known, so how does this works. Because now the equasion is: a = b-c/c. Where in you don't know a and b?

And the other question why do you fillna two times?

[amaboh]

minor is no longer supporter in np.range() function

[rubcaspac]

I am having a very stupid problem, come list have the share type using a .B or .A type, others have -B and -A. What is a good strategy to avoid this conflict, I having a very recurrent problem with this. Thank you

[BrandonJacobson]

I just ran this (9 Sep 19). I wasn't able to use tickers = df.columns.values.tolist() as it says .values doesn't have .tolist() I'm hoping this doesn't impact the next video.

[zes]

not pronx by that, no such thing as typox, type anyx and anyx can be perfx. type and talk can be perfx

[robsonvonbrum]

I wanna hit the fucking monitor everytime he sings a word