Normalization and Clustering Flow
Node-RED flow for performing data normalization and clustering. It takes input data for normalization and clustering, applies the specified normalization technique (such as Min-Max or Mean), performs clustering using the K-means algorithm, and returns the cluster centroids and statistical properties of the input data in JSON format.
[{"id":"0bdfbac1f0c218e0","type":"tab","label":"Normalization Extension","disabled":false,"info":"","env":[]},{"id":"60e707bae46e2331","type":"http in","z":"0bdfbac1f0c218e0","name":"","url":"/run","method":"post","upload":false,"swaggerDoc":"","x":400,"y":220,"wires":[["61b34aabae37d071"]]},{"id":"22fcc7626accc6f6","type":"http response","z":"0bdfbac1f0c218e0","name":"","statusCode":"","headers":{},"x":950,"y":220,"wires":[]},{"id":"49d8a6eef4cb19ab","type":"debug","z":"0bdfbac1f0c218e0","name":"debug 5","active":true,"tosidebar":true,"console":false,"tostatus":false,"complete":"true","targetType":"full","statusVal":"","statusType":"auto","x":940,"y":140,"wires":[]},{"id":"61b34aabae37d071","type":"function","z":"0bdfbac1f0c218e0","name":"Clustering function","func":"const clusterMaker = global.get('clusters');\nvar centroids = [];\nlet normalizationType = \"None\";\nlet data_statistical_properties; \n\n// Retrieve the parameters from the input message or use default values\nconst numClusters = msg.payload.value.numClusters || 3;\nconst numIterations = msg.payload.value.numIterations || 750;\n\nclusterMaker.k(numClusters);\nclusterMaker.iterations(numIterations);\n\n// Function to perform Min-Max normalization\nfunction minMaxNormalization(data, clustering_type) {\n if (clustering_type == \"multiple\"){\n const min = Math.min(...data);\n const max = Math.max(...data);\n const scaledData = data.map(value => [(value[0] - min) / (max - min)]);\n return scaledData;\n } \n else if (clustering_type == \"single\"){\n const min = Math.min(...data);\n const max = Math.max(...data);\n const scaledData = data.map(value => (value - min) / (max - min));\n return scaledData;\n } \n}\n\n// Function to perform mean normalization\nfunction meanNormalization(data, clustering_type) {\n if (clustering_type == \"multiple\") {\n const values = data.map(value => value[0]);\n const mean = values.reduce((sum, value) => sum + value, 0) / values.length;\n const stdDev = Math.sqrt(\n values.reduce((sum, value) => sum + Math.pow(value - mean, 2), 0) / values.length\n );\n const scaledData = data.map(value => [(value[0] - mean) / stdDev]);\n return scaledData;\n }\n else if (clustering_type == \"single\") {\n const sum = data.reduce((acc, value) => acc + value, 0);\n const mean = sum / data.length;\n const stdDev = Math.sqrt(\n data.reduce((sum, value) => sum + Math.pow(value - mean, 2), 0) / data.length\n );\n const scaledData = data.map(value => (value - mean) / stdDev);\n return scaledData;\n } \n}\n\nfunction get_data_statistical_properties(data){\n\n const stat_properties = [];\n\n for (const key in data) {\n const resource = data[key];\n const valueArray = resource.value;\n const min = Math.min(...valueArray);\n const max = Math.max(...valueArray);\n const sum = valueArray.reduce((acc, value) => acc + value, 0);\n const mean = sum / valueArray.length;\n const stdDev = Math.sqrt(\n valueArray.reduce((sum, value) => sum + Math.pow(value - mean, 2), 0) / valueArray.length\n );\n stat_properties.push({\n name: resource.name,\n value: {\"min\": min, \"max\":max, \"mean\":mean, \"stdDev\":stdDev}\n });\n }\n\n return stat_properties;\n}\n\n\nif (msg.payload.value.mode === \"multiple\") {\n // For each resource\n msg.payload.value.data.forEach(element => {\n // Creating clusterMakerInput\n let clusterMakerInput = [];\n element.value.forEach(value => {\n clusterMakerInput.push([value]);\n });\n\n // Print the unnormalized data\n //node.warn(\"Unnormalized data:\");\n //node.warn(clusterMakerInput);\n\n // Apply feature scaling based on the user-specified normalization technique\n if (msg.payload.value.normalization === \"minmax\") {\n //clusterMakerInput = clusterMakerInput.map(data => minMaxNormalization(data));\n clusterMakerInput = minMaxNormalization(clusterMakerInput, msg.payload.value.mode)\n normalizationType = \"MinMax\";\n } else if (msg.payload.value.normalization === \"mean\") {\n //clusterMakerInput = clusterMakerInput.map(data => meanNormalization(data));\n clusterMakerInput = meanNormalization(clusterMakerInput, msg.payload.value.mode)\n normalizationType = \"Mean\";\n }\n\n // Print the normalized data\n //node.warn(\"Normalized data:\");\n //node.warn(clusterMakerInput);\n\n clusterMaker.data(clusterMakerInput);\n\n let temp = clusterMaker.clusters();\n\n // Creating the output\n let clusterMakerOutput = [];\n temp.forEach(element => {\n element.centroid.forEach(element2 => {\n clusterMakerOutput.push(element2);\n });\n });\n clusterMakerOutput.sort();\n centroids.push({ [element.name]: clusterMakerOutput });\n });\n} else if (msg.payload.value.mode === \"single\") {\n\n let input_values;\n const normalizedData = [];\n\n // Apply feature scaling based on the user-specified normalization technique\n if (msg.payload.value.normalization === \"minmax\") {\n for (const key in msg.payload.value.data) {\n const resource = msg.payload.value.data[key];\n const valueArray = resource.value;\n const normalizedArray = minMaxNormalization(valueArray, \"single\");\n normalizedData.push({\n name: resource.name,\n value: normalizedArray\n });\n }\n input_values = normalizedData;\n normalizationType = \"MinMax\";\n } else if (msg.payload.value.normalization === \"mean\") {\n for (const key in msg.payload.value.data) {\n const resource = msg.payload.value.data[key];\n const valueArray = resource.value;\n const normalizedArray = meanNormalization(valueArray, \"single\");\n normalizedData[key] = {\n name: resource.name,\n value: normalizedArray\n };\n }\n\n input_values = normalizedData;\n //node.warn(input_values);\n normalizationType = \"Mean\";\n }\n else {\n input_values = msg.payload.value.data;\n //node.warn(input_values);\n normalizationType = \"None\";\n }\n\n // data.length = m\n // data.element1.length = n\n // Finding the maximum length of all data arrays\n let n = [];\n\n for (let m = 0; m < input_values.length; m++) {\n n.push(input_values[m].value.length);\n }\n\n const nMax = Math.max(...n);\n\n // Creating the array for clusterMakerInput\n let clusterMakerInput = [];\n for (let n = 0; n < nMax; n++) {\n let arrTemp = [];\n input_values.forEach(m => {\n arrTemp.push(m.value[n]);\n });\n node.warn(arrTemp);\n clusterMakerInput.push(arrTemp);\n }\n\n clusterMaker.data(clusterMakerInput);\n centroids = clusterMaker.clusters();\n\n}\n\ndata_statistical_properties = get_data_statistical_properties(msg.payload.value.data)\n\nmsg.payload = { centroids, normalizationType, data_statistical_properties};\nreturn msg;","outputs":1,"noerr":0,"initialize":"","finalize":"","libs":[],"x":730,"y":220,"wires":[["22fcc7626accc6f6","49d8a6eef4cb19ab"]]},{"id":"3bb826503eae074c","type":"http in","z":"0bdfbac1f0c218e0","name":"","url":"/init","method":"post","upload":false,"swaggerDoc":"","x":540,"y":400,"wires":[["386d1d1a05338043"]]},{"id":"386d1d1a05338043","type":"http response","z":"0bdfbac1f0c218e0","name":"","statusCode":"","headers":{},"x":710,"y":400,"wires":[]},{"id":"326aa812e0eb4ed0","type":"catch","z":"0bdfbac1f0c218e0","name":"","scope":null,"uncaught":false,"x":540,"y":320,"wires":[["e8e606dc288a1422"]]},{"id":"e8e606dc288a1422","type":"function","z":"0bdfbac1f0c218e0","name":"ADD ERROR INFO","func":"var payload=msg.payload;\nmsg.payload={};\n\nmsg.payload.error=msg.error;\nmsg.payload.error.payload=payload;\nreturn msg;","outputs":1,"noerr":0,"initialize":"","finalize":"","libs":[],"x":750,"y":320,"wires":[["22fcc7626accc6f6"]]},{"id":"dee65f181f179fbf","type":"inject","z":"0bdfbac1f0c218e0","name":"Inject","props":[{"p":"payload"},{"p":"topic","vt":"str"}],"repeat":"","crontab":"","once":false,"onceDelay":0.1,"topic":"","payload":"","payloadType":"date","x":390,"y":140,"wires":[["a3c8e2c4420b78c8"]]},{"id":"a3c8e2c4420b78c8","type":"function","z":"0bdfbac1f0c218e0","name":"Test input","func":"msg.payload = {\n \"value\": {\n \"mode\": \"single\",\n \"normalization\": \"mean\",\n \"data\": [\n {\n \"name\": \"Resource1\",\n \"value\": [1, 2, 3, 4, 5]\n },\n {\n \"name\": \"Resource2\",\n \"value\": [6, 7, 8, 9, 15]\n }\n ],\n \"numClusters\": 3,\n \"numIterations\": 1000\n }\n};\nreturn msg;\n","outputs":1,"noerr":0,"initialize":"","finalize":"","libs":[],"x":540,"y":140,"wires":[["61b34aabae37d071"]]}]
