Problem description: I have a pandas DataFrame with a "statistical" column that contains probabilities expressed as values ranging between 0 and 1. I would like to translate these values into discrete outputs as follows: {0.0 < x < 0.3 => 0, 0.7 < x < 1.0 => 1, 0.3 < x < 0.7 => 9}.
This is the piece of code I have so far. It does the job, but I believe it could be written in a better way.
df_predictions["discrete"] = df_predictions["statistical"]
df_predictions["discrete"][df_predictions["statistical"] < 0.3] = 0
df_predictions["discrete"][df_predictions["statistical"] > 0.7] = -1
df_predictions["discrete"][df_predictions["statistical"] > 0.1] = 9
df_predictions["discrete"][df_predictions["statistical"] < -0.1] = 1
Intuitively, it seems like the pandas.map method could be appropriate for this task, but I'm having trouble getting it to work. Maybe it's not suitable.
df_predictions["discrete"] = df_predictions["discrete"].map([{x < 0.3:1, x > 0.7:1, 0.3 < x < 0.7:9} for x in df_predictions["discrete"]])
I also tried numpy.select method, but without much success.
df_predictions["discrete"] = np.select([df_predictions["discrete"] < 0.3, df_predictions["discrete"] > 0.7, 0.3 < df_predictions["discrete"] < 0.7], [0, 1, 9])
I would appreciate any assistance with this. Looking for a simple solution, preferably one-liner.