problem of anomaly score when using HTM in anomaly detection

[Jianghao Shen]

hello everyone, I've been trying to apply HTM in detecting anomalies in ECG
data, i have put the result in the attachment, where the yellow line is the
anomaly score , the pink one represent the prediction , blue is the actual
data. My problem is that ideally the anomaly score should only be high in
the region where the data showed a unprecedented rising up and then drop to
a new low point ,and close to zero in other areas . as can be seen from the
picture, but the actual anomaly score are a series of discrete high value
distributed in the whole image ,with a more concentrated high score in the
region of anomalous , is there any way to fix this ,  clearer anomaly
score?

this is the json description file I wrote for the data:
 {
    "includedFields": [
        {
            "fieldName": "arythmia2",
            "fieldType": "float",
            "maxValue": 1.105,
            "minValue": -0.945
            }
        ],
    "streamDef": {
        "info": "arythmia2",
        "version": 1,
        "streams": [
            {
                "info": "arythmia2.csv",
                "source": "file://arythmia2.csv",
                "columns": [
                    "*"
                    ]
                }
            ]
        },
    "inferenceType": "TemporalAnomaly",
    "inferenceArgs": {
        "predictionSteps": [
            5
            ],
        "predictedField": "arythmia2"
        },
    "swarmSize": "medium"
    }


and after feed it to the swarm ,I then use the run_opf_experiement function
directly to the model paramters.