Machine Learning Program

Question

I've written a program that finds the difference between data and gives output. Here's the code:

import json
import numpy as np
print("This is a basic machine learning thing.")
baseData = {"collecting":True,"x":[],"y":[]}
while baseData["collecting"]:
  baseData["x"].append(float(input("X:")))
  baseData["y"].append(float(input("Y:")))
  if input("Do you want to keep feeding data? Press enter for yes, or type anything for no.") != "":
    baseData["collecting"] = False
if len(baseData["x"]) == len(baseData["y"]):
  xdata = baseData["x"]
  ydata = baseData["y"]
  nums = []
  for i in range(len(xdata)):
    nums.append(xdata[i] - ydata[i])
  median = np.median(nums)
else:
  print("malformed data")
def getY(x):
  pass
while True:
  data = input("X/Data:")
  print(int(data)-median)

To work the program, give it X and Y data, then give it X data and it will predict Y data.

Answer 1

Maybe you should check validity of the input and ask again if input is wrong?

while baseData["collecting"]:
  baseData["x"].append(float(input("X:")))

This is always True, so just discard this part:

if len(baseData["x"]) == len(baseData["y"]):

Maybe you should give an option to exit?

while True:
  data = input("X/Data:")
  print(int(data)-median)

And in general, calling this a "machiene-learning thing" is quite a stretch of imagination, no? Maybe you should have a look at how to fit data. I.e. a basic linear model fit.

Answer 2

baseData should be split into 3 separate variables(collecting, xdata, ydata); there is no reason for it to be a dict.

nums = []
for i in range(len(xdata)):
  nums.append(xdata[i] - ydata[i])

can be written more Pythonically as:

nums = []
for x, y in zip(xdata, ydata):
  nums.append(x - y)

or even just:

nums = [x - y for x, y in zip(xdata, ydata)]

You don't need to import numpy only for the media; stdlib statistics.median should work just fine.