resampling a pandas dataframe and filling new rows with zero - python-3.x

I have a time series as a dataframe. The first column is the week number, the second are values for that week. The first week (22) and the last week (48), are the lower and upper bounds of the time series. Some weeks are missing, for example, there is no week 27 and 28. I would like to resample this series such that there are no missing weeks. Where a week was inserted, I would like the corresponding value to be zero. This is my data:
week value
0 22 1
1 23 2
2 24 2
3 25 3
4 26 2
5 29 3
6 30 3
7 31 3
8 32 7
9 33 4
10 34 5
11 35 4
12 36 2
13 37 3
14 38 10
15 39 5
16 40 7
17 41 10
18 42 11
19 43 15
20 44 9
21 45 13
22 46 5
23 47 6
24 48 2
I am wondering if this can be achieved in Pandas without creating a loop from scratch. I have looked into pd.resample, but can't achieve the results I am looking for.

I would set week as index, reindex with fill_value option:
start, end = df['week'].agg(['min','max'])
df.set_index('week').reindex(np.arange(start, end+1), fill_value=0).reset_index()
Output (head):
week value
0 22 1
1 23 2
2 24 2
3 25 3
4 26 2
5 27 0
6 28 0
7 29 3
8 30 3

Related

How to insert a pandas series as a new column in DataFrame, matching with the indexes of df with series of different length

I have a dataframe with multiple columns and 700+ rows and a series of 27 rows. I want to create a new column i.e. series in dataframe as per matching indexes with predefined column in df
data frame I have and need to add series which contains the same indexes of "Reason for absence"
ID Reason for absence Month of absence Day of the week Seasons
0 11 26 7 3 1
1 36 0 7 3 1
2 3 23 7 4 1
3 7 7 7 5 1
4 11 23 7 5 1
5 3 23 7 6 1
6 10 22 7 6 1
7 20 23 7 6 1
8 14 19 7 2 1
9 1 22 7 2 1
10 20 1 7 2 1
11 20 1 7 3 1
12 20 11 7 4 1
13 3 11 7 4 1
14 3 23 7 4 1
15 24 14 7 6 1
16 3 23 7 6 1
17 3 21 7 2 1
18 6 11 7 5 1
19 33 23 8 4 1
20 18 10 8 4 1
21 3 11 8 2 1
22 10 13 8 2 1
23 20 28 8 6 1
24 11 18 8 2 1
25 10 25 8 2 1
26 11 23 8 3 1
27 30 28 8 4 1
28 11 18 8 4 1
29 3 23 8 6 1
30 3 18 8 2 1
31 2 18 8 5 1
32 1 23 8 5 1
33 2 18 8 2 1
34 3 23 8 2 1
35 10 23 8 2 1
36 11 24 8 3 1
37 19 11 8 5 1
38 2 28 8 6 1
39 20 23 8 6 1
40 27 23 9 3 1
41 34 23 9 2 1
42 3 23 9 3 1
43 5 19 9 3 1
44 14 23 9 4 1
this is series table s_conditions
0 Not absent
1 Infectious and parasitic diseases
2 Neoplasms
3 Diseases of the blood
4 Endocrine, nutritional and metabolic diseases
5 Mental and behavioural disorders
6 Diseases of the nervous system
7 Diseases of the eye
8 Diseases of the ear
9 Diseases of the circulatory system
10 Diseases of the respiratory system
11 Diseases of the digestive system
12 Diseases of the skin
13 Diseases of the musculoskeletal system
14 Diseases of the genitourinary system
15 Pregnancy and childbirth
16 Conditions from perinatal period
17 Congenital malformations
18 Symptoms not elsewhere classified
19 Injury
20 External causes
21 Factors influencing health status
22 Patient follow-up
23 Medical consultation
24 Blood donation
25 Laboratory examination
26 Unjustified absence
27 Physiotherapy
28 Dental consultation
dtype: object
I tried this
df1.insert(loc=0, column="Reason_for_absence", value=s_conditons)
out- this is wrong because i need the reason_for_absence colum according to the index of reason for absence and s_conditions
Reason_for_absence ID Reason for absence \
0 Not absent 11 26
1 Infectious and parasitic diseases 36 0
2 Neoplasms 3 23
3 Diseases of the blood 7 7
4 Endocrine, nutritional and metabolic diseases 11 23
5 Mental and behavioural disorders 3 23
6 Diseases of the nervous system 10 22
7 Diseases of the eye 20 23
8 Diseases of the ear 14 19
9 Diseases of the circulatory system 1 22
10 Diseases of the respiratory system 20 1
11 Diseases of the digestive system 20 1
12 Diseases of the skin 20 11
13 Diseases of the musculoskeletal system 3 11
14 Diseases of the genitourinary system 3 23
15 Pregnancy and childbirth 24 14
16 Conditions from perinatal period 3 23
17 Congenital malformations 3 21
18 Symptoms not elsewhere classified 6 11
19 Injury 33 23
20 External causes 18 10
21 Factors influencing health status 3 11
22 Patient follow-up 10 13
23 Medical consultation 20 28
24 Blood donation 11 18
25 Laboratory examination 10 25
26 Unjustified absence 11 23
27 Physiotherapy 30 28
28 Dental consultation 11 18
29 NaN 3 23
30 NaN 3 18
31 NaN 2 18
32 NaN 1 23
i am getting output upto 28 rows and NaN values after that. Instead, I need correct order of series according to indexes for all the rows
While this question is a bit confusing, it seems the desire is to match the series index with the dataframe "Reason for Absence" column. If this is correct, below is a small example of how to accomplish. Keep in mind, the resulting dataframe will be sorted based on the 'Reason for Absence Numerical' column. If my understanding is incorrect, please clarify this question so we can better assist you.
d = {'ID': [11,36,3], 'Reason for Absence Numerical': [3,2,1], 'Day of the Week': [4,2,6]}
dataframe = pd.DataFrame(data=d)
s = {0: 'Not absent', 1:'Neoplasms', 2:'Injury', 3:'Diseases of the eye'}
disease_series = pd.Series(data=s)
def add_series_to_df(df, series, index_val):
df_filtered = df[df['Reason for Absence Numerical'] == index_val].copy()
series_filtered = series[series.index == index_val]
if not df_filtered.empty:
df_filtered['Reason for Absence Text'] = series_filtered.item()
return df_filtered
x = [add_series_to_df(dataframe, disease_series, index_val) for index_val in range(len(disease_series.index))]
new_df = pd.concat(x)
print(new_df)

Sort rows by row value (top to bottom)

There is lotto draw (5 numbers) on each row. I have formula which calculates the most frequient numbers with their number of draws. Is it possible in end result to sort same number of draws results by row value. This means that if number is drawn on top rows will have grater value than those on bottom rows. Considering number of row to be a value. How is that possible?
Formula used:
=LET(flatten, TEXTSPLIT(TEXTJOIN(";",,A1:F27),,";"), numUq, UNIQUE(flatten), matches, XMATCH(flatten,numUq),SORT(HSTACK(numUq, DROP(FREQUENCY(matches, UNIQUE(matches)),-1)),2,-1))
In the example screenshot number 35 and number 13 have equal draws count, but 13 should be before 35.
Data:
A
B
C
D
E
F
18
35
31
13
37
10
43
47
36
13
6
19
6
12
6
35
14
1
43
24
45
7
21
16
37
39
44
24
12
40
39
8
34
28
49
46
27
44
15
46
45
12
22
0
10
5
28
28
4
7
23
6
44
41
30
22
47
13
29
29
37
9
26
44
39
10
30
17
21
20
41
22
43
35
0
22
13
9
14
22
42
20
32
21
13
38
48
6
14
2
11
47
20
20
23
6
22
26
1
25
45
31
27
39
6
44
3
24
22
45
34
17
5
13
16
23
20
7
30
16
25
21
7
34
1
35
32
34
1
9
10
32
23
35
11
3
6
12
5
30
4
20
33
15
26
10
8
28
16
11
21
14
3
38
10
42
16
3
26
48
30
28
Link to file
Here it is on a bit of the data. Here I have added a third column based on the average row of each unique number and sorted first on frequency then on row average:
=LET(range,A1:F3,uniques,UNIQUE(TOCOL(range)),rows,SEQUENCE(ROWS(range)),
avrow,BYROW(uniques,LAMBDA(uniq,SUM((range=uniq)*rows/SUM(--(range=uniq))))),
freq,DROP(FREQUENCY(range,uniques),-1),
SORTBY(HSTACK(uniques,freq,avrow),freq,-1,avrow,1))
Can 6 really occur twice in the same draw? Maybe not, but it doesn't affect the answer.
EDIT
Here is a version based on your original formula:
=LET(range,A1:F27,
flatten, TEXTSPLIT(TEXTJOIN(";",,A1:F27),,";"),
numUq, UNIQUE(flatten),
rows,SEQUENCE(ROWS(range)),
matches, XMATCH(flatten,numUq),
avrow,BYROW(numUq,LAMBDA(numUq,SUM((range=--numUq)*rows/SUM(--(range=--numUq))))),
freq,DROP(FREQUENCY(matches, UNIQUE(matches)),-1),
SORTBY(HSTACK(numUq,freq,avrow),freq,-1,avrow,1))
Full Dataset
The sorting is based on number of appearances and average row, but you could use other measures like row of first appearance if you wanted to.
Different approach:
=LET(data,A1:F27,
a,TOCOL(data),
b,MMULT(--(TRANSPOSE(a)=a),SEQUENCE(COUNTA(a),,1,0)),
c,TOCOL(IF(ISNUMBER(data),MAX(ROW(data)+1)-ROW(data)^99)),
d,MMULT(--(TRANSPOSE(a)=a),c),
s,SORTBY(HSTACK(a,b),b,-1,d,1),
UNIQUE(s))
a "flattens" the data using TOCOL.
b creates a "countif" of the drawn values in a using MMULT.
c returns the maximum row value of the data + 1 minus the row value of each value found ^99.
^99 because I want the number to be higher if it would be found in the first row only versus if it was found in each row except the first.
d returns a "sumif" of the calculated row values of c against the values of a.
We than only need a and b for the list using HSTACK, but we need them sorted by the count b descending and sorted by the sumif d ascending using SORTBY.
This will sort it as you illustrated it.
If it's a tie (36 and 19 in the data) it will show the first in row first.

Normalizing the data using python

I have the below data
df1
Hema shiva Ishan
0 22 30 33
1 34 32 21
2 20 12 14
3 26 14 18
4 12 28 17
5 30 11 22
6 18 15 18
7 19 18 19
8 22 20 32
I wanted to take ratio of first column value with rest of the columns , eg first column should divide by 22 , 2nd column 30 and 3rd columns by 33 .
The answer is below .
Please help me if I missing something
Just divide the first row by the DF:
df.iloc[0] / df

Excel Pivot table - get maximum for a period of 24 hours

I have an excel with:
Days of the week and 24 hours for each day.
Each hour I get some points.
I would like to calcute the maximum of cumulate points I can get within 24 hours.
[TEST.XLSX]
2 Columns:
Monday Points
0 34
1 32
2 4
3 54
4 12
5 55
6 4
7 4
8 555
9 787
10 8
11 76
12 78
13 8
14 656
15 7
16 4
17 45
18 54
19 543
20 56
21 65
22 4
23 3
Tuesday
0 56
1 7
2 333
3 9
4 876
5 3333
6 3333
7 76
8 3333
9 465
10 7
11 6
12 5
13 6
14 7
15 6
16 7
17 65
18 555555555
19 6
20 5
21 4
22 6
23 6
Wednesday
0 6
1 7
...
Thanks for your help!
Use real date time values in your hours column. Delete the rows with the day text. Instead, use a formula that increments from a starting date/time. For example: cell A2 contains the date and midnight time for Nov 17. Cell A3 and copied down contains the formula
=A2+TIME(1,0,0)
which increments by one hour.
Now you con build a pivot table. Group by the date/time value by day and hour. Show the subtotal for the day and set its value field settings to Max.

pulling out the result from MATLAB to Excel row by row

I do calculations on 64 elements (for p=1:64 function end) and pull out the result values in an Excel file.
Is there any way to arrange the result values for each element row by row (the values of the first element should appear on the first row, the values of the second element should appear on the second row and so on)?
I used P=reshape(A,[],16) but Matlab pushes the values from right to the left mixing them.
For example,
If I set the loop for the calculation p=1:1 and use P=reshape(A,[],16) the result is:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
If I set p=1:2 the result becomes:
for element 1: 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
for element 2: 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
(the values of element 2 are: 17 18 19 20 21 22 23 24 25 ... 32)
The result for p=1:2 should be:
for element 1: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
for element 2: 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
for element 3: 33 34 35 ,etc...
Try this:
P=reshape(A,16,[])'
Is this what you need?

Resources