我想根据时间信息计算项目(0,1,2,3….)之间的相似性。时间信息可以是时间即时(startdate)、时间间隔(startdate、enddate)或null(NaT)；请参阅下面的dataframe(df_for)的示例。在

{instant，instant}：如果等于sim=1，则sim=0

{instant，null}或相反，sim=0

{instant，interval}：如果instant在interval内，sim=1，或者如果一个interval包含一个instant，sim=1

{interval，interval}：如果区间重叠，sim=两个区间的交集/两个区间的并集

{interval，interval}：如果一个间隔包含另一个，则sim=1

下面的python代码从dataframe获取时态信息，并执行上述条件(1-5)。代码很冗长，我想知道是否有一种聪明的方法/lib来使用python计算时间段和时间瞬间之间的相似性。在m, k = df_for.shape

sim = np.zeros((m, m))

data = df_for.values

for i in range(m):

for j in range(m):

if i != j:

st1 = data[i][0]

ed1 = data[i][1]

st2 = data[j][0]

ed2 = data[j][1]

#both items are null values

if pd.isnull(st1) and pd.isnull(ed1) and pd.isnull(st2) and pd.isnull(ed2):

sim[i][j] = 0.

# {instant, instant} => equal, not equal

if pd.notnull(st1) and pd.isnull(ed1) and pd.notnull(st2) and pd.isnull(ed2):

if st1 == st2:

sim[i][j] = 1.

else:

sim[i][j] = 0.

# {instant, null} => sim is 0

if pd.notnull(st1) and pd.isnull(ed1) and pd.isnull(st2) and pd.isnull(ed2):

sim[i][j] = 0.

# {instant, interval} => meets, during

if pd.notnull(st1) and pd.isnull(ed1) and pd.notnull(st2) and pd.notnull(ed2):

if(st2 <= st1 <= ed2):

sim[i][j] = 1. #a time is between two other times

else:

sim[i][j] = 0.

# {interval, instant} => meets, contains

if pd.notnull(st1) and pd.notnull(ed1) and pd.notnull(st2) and pd.isnull(ed2):

if(st1 <= st2 <= ed1):

sim[i][j] = 1. #a time is between two other times

else:

sim[i][j] = 0.

# {interval, interval} => equal, overlaps, not overlaps

if pd.notnull(st1) and pd.notnull(ed1) and pd.notnull(st2) and pd.notnull(ed2):

if (st1 <= st2 <= ed1) or (st2 <= st1 <= ed2):

intersect = min(ed1,ed2)- max(st1,st2) # earliestend-lateststart

union = max(st1,st2,ed1,ed2) - min(ed1,ed2,st1,st2)

overlaps = intersect/union

#print(intersect/np.timedelta64(1, 'D'),union/np.timedelta64(1, 'D'))

if (st1 > st2 and ed1 < ed2) or (st1 < st2 and ed1 > ed2): # contains, during

overlaps = 1.0

sim[i][j]=overlaps

else:

sim[i][j] = 0.

else:

sim[i][j] = 1.