[FIXED] Pandas - Group by Cumulative Calculation

Issue

I'm trying to do a calculation with cumulative operations using pandas in jupyter notebook.

But I'm stuck.

I have the following DataFrame:

Date	Type	deposit	withdrawal
2020/01/19	A	1000	0
2020/01/23	A	500	0
2020/02/05	A	0	200
2020/02/22	A	1000	200
2020/01/19	B	859.8	0
2020/02/04	B	514.33	50
2020/02/25	B	1463.14	0

I need to apply this formula, grouping by type:

Movement = IF(ISNULL([previous_date]); [Current_Deposit] - [withdrawal];([previous_movement] * (1 +([current_date]-[previous_date]) * 0,0001)) + [Current_Deposit] - [withdrawal])

Example:

Date	Type	deposit	withdrawal	Movement	Formula
2020/01/19	A	1000	0	1000	IF(ISNULL([NULL]); [1000] - [0];([0] * (1 +([2020/01/19]-[NULL]) * 0,0001)) + [1000] - [0])
2020/01/23	A	500	0	1500.40	IF(ISNULL([2020/01/19]); [500] - [0];([1000] * (1 +([2020/01/23]-[2020/01/19]) * 0,0001)) + [500] - [0])
2020/02/05	A	0	200	1302.35	IF(ISNULL([2020/01/23]); [0] - [200];([1500.40] * (1 +([2020/02/05]-[2020/01/23]) * 0,0001)) + [0] - [200])
2020/02/22	A	1000	200	2104,56	IF(ISNULL([2020/02/05]); [1000] - [200];([1302.35] * (1 +([2020/02/22]-[2020/02/05]) * 0,0001)) + [1000] - [200])
2020/01/19	B	859.8	0	859.80	IF(ISNULL([NULL]); [859.8] - [0];([0] * (1 +([2020/01/19]-[NULL]) * 0,0001)) + [859.8] - [0])
2020/02/04	B	514.33	50	1325.51	IF(ISNULL([2020/01/19]); [514.33] - [50];([859.80] * (1 +([2020/02/04]-[2020/01/19]) * 0,0001)) + [514.33] - [50])
2020/02/25	B	1463.14	0	2791.43	IF(ISNULL([2020/02/04]); [1463.14] - [0];([1325.51] * (1 +([2020/02/25]-[2020/02/04]) * 0,0001)) + [1463.14] - [0])

I'm not able to accumulate the values to put in the formula.

I need to group the calculations by type, and accumulate the values to use in the current line. I'm new to python and I'm not able to implement it.

Solution

I assume that your dataframe is sorted by Type and Date

# silencing chained assignment warning
pd.options.mode.chained_assignment = None

df['Movement'] = 0

df_new = df
for row in df.iterrows():
    x = row[1]
    y = df_new.shift().loc[x.name]
    if x['Type'] == y['Type']:
        # computing the given days compound interest should calculate using power
        compound_interest = pow(1.0001, pd.Timedelta(pd.to_datetime(x['Date']) - pd.to_datetime(y['Date'])).days)
        # if you wish to use your formula uncomment the following line
        # compound_interest = 1 + 0.0001 * pd.Timedelta(pd.to_datetime(x['Date']) - pd.to_datetime(y['Date'])).days
        df_new['Movement'].loc[x.name] = compound_interest * y['Movement'] + x['deposit'] - x['withdrawal']
    else:
        df_new['Movement'].loc[x.name] = x['deposit']

print(df_new)

OUTPUT

      Date Type  deposit  withdrawal    Movement
2020/01/19    A  1000.00           0 1000.000000
2020/01/23    A   500.00           0 1500.400060
2020/02/05    A     0.00         200 1302.351751
2020/02/22    A  1000.00         200 2104.567521
2020/01/19    B   859.80           0  859.800000
2020/02/04    B   514.33          50 1325.506712
2020/02/25    B  1463.14           0 2791.433062

Answered By - Ze'ev Ben-Tsvi