search

ameybarapatre / ANFIS

Fuzzy Q learning in Tensorflow , Training an ANFIS.
MIT License
25 stars 5 forks source link

pandas update #3

Open BotYue opened 6 years ago

BotYue commented 6 years ago

Hi There,

I tried to reproduce your work but got some trouble when runing 'run.py'. I modified from line 11.

data['Diff'] = data['Diff']*15
data['Q'] = np.ravel(E)
output = data.sort_values(['Q'] , ascending=False)
output.to_csv('output.csv')
data[['Diff' ,'Q']].plot()
plt.show()

I think this is caused by some issues and updates from pandas. My versions are python 2.7, pandas 0.20.3.

And what is the meaning of the final plot? How does it look? Thanks! figure_1

ameybarapatre commented 6 years ago

Hi ,

Firstly, sorry for getting back so late. Also, I have updated the Readme but the gist of the matter is 'Diff' is the improvement in the rate at which the taxi is making money if he takes a particular ride. Please look at preprocessing.py the column names of data are very suggestive of what it represents.

The reward is calculated as the change in the PID value if a taxi takes a fare. I'll soon attach a paper to explain the project.

Q in the plot is the output of the ANFIS, that is Q value of selecting a fare.

So the final plot shows the Diff value and the Q value of selecting a fare from the dataset where the x-axis is the row number in the data set.