PoSeiDon-Workflows / FlowGAD

Workflow Anomaly Detection with Graph Neural Networks
MIT License
0 stars 0 forks source link

Shift timestamp based on parent node #10

Open cshjin opened 1 year ago

cshjin commented 1 year ago

Timestamp columns:

'ready', 'submit', 'execute_start', 'execute_end', 'post_script_start', 'post_script_end'

In general, they have the following condition per node

$$ \text{ready} \le \text{submit} \le \text{execute start} \le \text{execute end} \le \text{post script start} \le \text{post script end} $$

cshjin commented 1 year ago

Two options:

# global min over graph-level timestamps
ts_anchor = df['ready'].min()
tfs = ['ready', 'submit', 'execute_start', 'execute_end', 'post_script_start', 'post_script_end']
for attr in tfs[::-1]:
    df[attr] -= ts_anchor