Skip to content
LQM Registry / monitoring

Two-Sided CUSUM Drift Detector

Sequential two-sided cumulative-sum detector for upward or downward shifts in realized error, residual, risk or other quantitative streams.

Model keylq.monitoring.two-sided-cusum
Version1.0.0
Runtimepython-process
Run governed modelTry browser demo
Model contract

Inputs are explicit.

{
    "type": "object",
    "required": [
        "values"
    ],
    "properties": {
        "values": {
            "type": "array",
            "items": {
                "type": "number"
            },
            "minItems": 8
        }
    }
}
Output contract

Outputs are explicit.

{
    "type": "object",
    "required": [
        "drift_detected",
        "alarms",
        "reference",
        "latest_statistic",
        "max_statistic",
        "sequential_guard"
    ]
}
Default parameters
{
    "warmup": 8,
    "slack": 0,
    "threshold": 1
}
Example input
{
    "values": [
        0.1000000000000000055511151231257827021181583404541015625,
        0.11000000000000000055511151231257827021181583404541015625,
        0.0899999999999999966693309261245303787291049957275390625,
        0.1000000000000000055511151231257827021181583404541015625,
        0.11999999999999999555910790149937383830547332763671875,
        0.11000000000000000055511151231257827021181583404541015625,
        0.1000000000000000055511151231257827021181583404541015625,
        0.0899999999999999966693309261245303787291049957275390625,
        0.450000000000000011102230246251565404236316680908203125,
        0.5,
        0.479999999999999982236431605997495353221893310546875,
        0.5300000000000000266453525910037569701671600341796875,
        0.4899999999999999911182158029987476766109466552734375,
        0.5500000000000000444089209850062616169452667236328125
    ]
}
Capabilities
cpumonitoringdriftstreamingsequential
← Back to public registry