CODEWITHSUNDEEP
pythonpython-3.x
LewisRa
LewisRa

Reputation: 21

TypeError: unsupported operand type(s) for +: 'int' and 'str' (Logistic regression)

When I run this predict code in a juypter notebook it runs perfectly and I tested it by predict([1,2,1,0,1], [.4,1,3,.01,.1]) in a separate file, I get the correct answer of0.995929862284 but when the use the unit test I receive the error below.

def dot(X, Y):
   if len(X) != len(Y):
      return 0

   return sum(i[0] * i[1] for i in zip(X, Y))

def predict(features, weights):
  x = dot(features, weights)
  return logistic(x)

def test_predict(self):
    model = [1,2,1,0,1]
    point = {'features':[.4,1,3,.01,.1], 'label': 1}
    p = predict(model, point)
    self.assertAlmostEqual(p, 0.995929862284)

Error:

enter image description here

Upvotes: 1

Views: 677

Answers (3)

Patrick Artner
Patrick Artner

Reputation: 51683

Great read: How to debug small programs (#2) - set a breakpoint inside your dot() sum and inspect what datatypes are in X && Y when using 

return sum(i[0] * i[1] for i in zip(X, Y)

when you supply a dict as Y to it in your test method.

Upvotes: 0

Jean-François Fabre
Jean-François Fabre

Reputation: 140307

your both codes aren't equivalent

point = {'features':[.4,1,3,.01,.1], 'label': 1}

you're passing a dictionary of lists instead of the value. When iterated upon, the dictionary yields strings. You mean:

p = predict(model, point['features'])

Upvotes: 2

chepner
chepner

Reputation: 532508

When you call predict, you pass a dict as the second argument. However, predict passes that argument as-is to dot, which expects a list of numbers instead. As a result, dot iterates over the keys of the dict, rather than the values of that dict's features value. predict should be called as predict(model, point['features']) instead. (Or perhaps predict(point['features'], model), given the parameter names.)

Upvotes: 2

Related Questions