alyakhtar / aqi-delhi Goto Github PK

@iCHAIT Readme to bande saaxy si!!

@alyakhtar -

• Remove all rows that have PM 2.5 value = 0
• Normalize input and output data
• Divide the data into training and testing data. Count the total rows, that count will be our total no. of observation points. Take 60% of these total observation points as training data. (Remember, take this 60% data as alternative days, as explained by Sarfaraz Sir)

• Confusion matrix for MLP
• Graph comparing PM 2.5 for the month of April for the last 4 years, similar to Jan_2016.png.
• Graph comparing both the odd-evens (just the 15 days).

Hi,
Thanks for publishing this code. However, I am not sure where to start and how to run this, can you please point me towards it? A few steps on getting started should be helpful.

Thanks.

• Find correlation between a combination of 2 inputs and output.
• Find data for gases such as CO2, CO, SO2, No?
• After finding data for gases apply mlp just as being done right now. (Essentially having gases provides more number of input factors.)
• Divide the data into training, testing and validation?
• Use Neuro-Fuzzy Networks. Use RNN instead?

• Plot 2 graphs using matplotlib -
• One showing that the loss is decreasing continuously - that is for the script mlp.py

Notice that the loss is decreasing

• Second graph involves plotting observed and expected PM 2.5 values, remember show both the curves on the same graph so that it is easier to compare.
• We have mean_sqaure_error now we need to calculate the accuracy for the algo's, for this we need to calculate True Positives (TP), False Positives (FP), True Negatives (TN), False Negatives (FP) and consequently create a confusion matrix. All this stuff should be in some python library just have a look at the docs.

Hey man,
I found the data for my city on cpcb website, but when i am clicking the get as csv file, it is stuck on an "info processing" notification. Have you faced a similar issue and how did you get your data? Did you use any bot for this?