def corr_analysis(data):
col = list(data.columns)
col.remove("CreationTime")
col.remove( "InBandwidth")
target_attribute = "InBandwidth"
cor = []
for i in col:
corval = abs(st.pearsonr(data[target_attribute],data[i])[0])
cor.append(corval)
return cor
def print_corr(data):
col = list(data.columns)
col.remove("CreationTime")
col.remove( "InBandwidth")
target_attribute = "InBandwidth"
cor = []
print("\nCorrelation Coefficient of 'InBandwidth' with ")
for i in col:
corval = abs(st.pearsonr(data[target_attribute],data[i])[0])
cor.append(corval)
print(i," = ",corval)
def corr_column(df):
col = list(df.columns)
#col.remove("CreationTime")
col.remove("InBandwidth")
val = []
for i in col:
val.append(abs(st.pearsonr(df["InBandwidth"], df[i])[0]))
x = sorted(val, reverse=True)
col1 = col[val.index(x[0])] #maximum correalatio coefficient
col2 = col[val.index(x[1])]
print(col1, col2)
return col1, col2
def corr_analysis(data):
col = list(data.columns)
#col.remove("CreationTime")
col.remove("InBandwidth")
target_attribute = "InBandwidth"
CORRs = []
print(
"\nCorrelation Coefficient of 'InBandwidth' with other attributes:\n")
for i in col:
cor = abs(st.pearsonr(data[target_attribute], data[i])[0])
CORRs.append(cor)
print(i, ": ", cor)
return CORRs
def loss_corr(x1, x2, y1, y2, criterion, model, corr_lambda):
loss1 = criterion(y1, x2)
loss2 = criterion(y2, x1)
corr_term = corr_lambda * stats.pearsonr(model.encoder1(x1),
model.encoder2(x2))
return loss1 + loss2 - corr_term