在 Python 中查找未知值

Question

I have a dataset with 500 restaurant orders, and their totals. I want to identify the outliers in the dataset. And then decide if they are valid data points or wrong values. And then remove the invalid ones.

The problem is I only have the total price of the orders, the item name, and the quantity ordered. I was wondering if it is possible to get the price of each item.

Each item in the dictionary presents one order. The Key is the total price, and the value is a list of tuples, each tuple present the item name and the quantity ordered.

Sample of my dataset in a dictionary format(I have it as two columns in a dataframe too):

{1215.5: [('Shrimp', 10), ('Fish&Chips', 6), ('Salmon', 8), ('Pasta', 5)],
 1230.0: [('Shrimp', 10), ('Salmon', 10), ('Fish&Chips', 8)],
 1234.0: [('Salmon', 9), ('Fish&Chips', 3), ('Pasta', 8), ('Shrimp', 10)],
 1292.5: [('Pasta', 7), ('Salmon', 9), ('Fish&Chips', 7), ('Shrimp', 9)],
 1301.5: [('Pasta', 5), ('Shrimp', 9), ('Salmon', 8), ('Fish&Chips', 10)],
 1314.5: [('Shrimp', 10), ('Pasta', 5), ('Fish&Chips', 10), ('Salmon', 7)],
 1343.5: [('Shrimp', 8), ('Fish&Chips', 10), ('Salmon', 9), ('Pasta', 7)]}

My desired output is to obtain the price of each item. By doing this I hope I can be able to decide if the total is a valid data point or an outlier.

I tried taking the third line and store the value in a list A

[('Salmon', 9), ('Fish&Chips', 3), ('Pasta', 8), ('Shrimp', 10)]

And the total price of these items B

[1234.0]

Then I tried converting my first list to an array

    A=np.array(lst)

The output

array([['Salmon', '9'],
       ['Fish&Chips', '3'],
       ['Pasta', '8'],
       ['Shrimp', '10']], dtype='<U10')

The shapes

A.shape
(4,2)
B.shape
(1,)

Then applied the function

X, _, _, _ = np.linalg.lstsq(A, B)

but the output returns an error message

LinAlgError: Incompatible dimensions

I know that m has to be equal for the function to work. But I am not sure how to change the shape of A .

Any input is appreciated. Thank you,

Answer 1

A possible solution would be to use construct a [possibly overdetermined] system of linear equations and solve it. For example, the first list becomes 1215.5=10*Shrimp+6*Fish+8*Salmon+5*Pasta .

Assuming that the name of your dictionary is d , the matrix A of the system is given by:

A = pd.concat([pd.DataFrame(d[x]).set_index(0) for x in d], axis=1)\
                 .fillna(0).T

(And I strongly suggest that you do not use dict as the storage container.) The vector B is the list of keys:

B = list(d.keys())

The answer is:

X, _, _, _ = numpy.linalg.lstsq(A, B)
#array([35. , 27.5, 41. , 54. ])

You only need the first part of the returned tuple.