Source Code for Module rdkit.ML.ModelPackage.PackageUtils

  1  # 
  2  # Copyright (C) 2003 Rational Discovery LLC 
  3  # All rights are reserved. 
  4  # 
  5  from elementtree.ElementTree import ElementTree,Element,SubElement 
  6  import time 
  7   
  8   
  9 -def _ConvertModelPerformance(perf,modelPerf): 
 10    if len(modelPerf)>3: 
 11      confMat = modelPerf[3] 
 12      accum = 0 
 13      for row in confMat: 
 14        for entry in row: 
 15          accum += entry 
 16      accum = str(accum)     
 17    else: 
 18      confMat = None 
 19      accum = 'N/A' 
 20   
 21    if len(modelPerf)>4: 
 22      elem = SubElement(perf,"ScreenThreshold") 
 23      elem.text=str(modelPerf[4]) 
 24    elem = SubElement(perf,"NumScreened") 
 25    elem.text=accum 
 26    if len(modelPerf)>4: 
 27      elem = SubElement(perf,"NumSkipped") 
 28      elem.text=str(modelPerf[6]) 
 29    elem = SubElement(perf,"Accuracy") 
 30    elem.text=str(modelPerf[0]) 
 31    elem = SubElement(perf,"AvgCorrectConf") 
 32    elem.text=str(modelPerf[1]) 
 33    elem = SubElement(perf,"AvgIncorrectConf") 
 34    elem.text=str(modelPerf[2]) 
 35    if len(modelPerf)>4: 
 36      elem = SubElement(perf,"AvgSkipConf") 
 37      elem.text=str(modelPerf[5]) 
 38    if confMat: 
 39      elem = SubElement(perf,"ConfusionMatrix") 
 40      elem.text = str(confMat) 
 41     
 42 -def PackageToXml(pkg,summary="N/A",trainingDataId='N/A', 
 43                   dataPerformance=[], 
 44                   recommendedThreshold=None, 
 45                   classDescriptions=[], 
 46                   modelType=None, 
 47                   modelOrganism=None): 
 48    """ generates XML for a package that follows the RD_Model.dtd 
 49   
 50    If provided, dataPerformance should be a sequence of 2-tuples: 
 51      ( note, performance ) 
 52    where performance is of the form: 
 53      ( accuracy, avgCorrectConf, avgIncorrectConf, confusionMatrix, thresh, avgSkipConf, nSkipped ) 
 54      the last four elements are optional   
 55   
 56    """ 
 57    head = Element("RDModelInfo") 
 58    name = SubElement(head,"ModelName") 
 59    notes = pkg.GetNotes() 
 60    if not notes: 
 61      notes = "Unnamed model" 
 62    name.text = notes 
 63    summ = SubElement(head,"ModelSummary") 
 64    summ.text = summary 
 65    calc = pkg.GetCalculator() 
 66    descrs = SubElement(head,"ModelDescriptors") 
 67    for name,summary,func in zip(calc.GetDescriptorNames(),calc.GetDescriptorSummaries(),calc.GetDescriptorFuncs()): 
 68      descr = SubElement(descrs,"Descriptor") 
 69      elem = SubElement(descr,"DescriptorName") 
 70      elem.text = name 
 71      elem = SubElement(descr,"DescriptorDetail") 
 72      elem.text = summary 
 73      if hasattr(func,'version'): 
 74        vers = SubElement(descr,"DescriptorVersion") 
 75        major,minor,patch = func.version.split('.') 
 76        elem = SubElement(vers,"VersionMajor") 
 77        elem.text = major 
 78        elem = SubElement(vers,"VersionMinor") 
 79        elem.text = minor 
 80        elem = SubElement(vers,"VersionPatch") 
 81        elem.text = patch 
 82   
 83    elem = SubElement(head,"TrainingDataId") 
 84    elem.text = trainingDataId 
 85   
 86    for description,perfData in dataPerformance: 
 87      dataNode = SubElement(head,"ValidationData") 
 88      note = SubElement(dataNode,'ScreenNote') 
 89      note.text = description 
 90      perf = SubElement(dataNode,"PerformanceData") 
 91      _ConvertModelPerformance(perf,perfData) 
 92   
 93   
 94    if recommendedThreshold: 
 95      elem = SubElement(head,"RecommendedThreshold") 
 96      elem.text=str(recommendedThreshold) 
 97   
 98    if classDescriptions: 
 99      elem = SubElement(head,"ClassDescriptions") 
100      for val,text in classDescriptions: 
101        descr = SubElement(elem,'ClassDescription') 
102        valElem = SubElement(descr,'ClassVal') 
103        valElem.text = str(val) 
104        valText = SubElement(descr,'ClassText') 
105        valText.text = str(text) 
106   
107    if modelType: 
108      elem = SubElement(head,"ModelType") 
109      elem.text=modelType 
110    if modelOrganism: 
111      elem = SubElement(head,"ModelOrganism") 
112      elem.text=modelOrganism 
113       
114       
115    hist = SubElement(head,"ModelHistory") 
116    revision = SubElement(hist,"Revision") 
117    tm = time.localtime() 
118    date = SubElement(revision,"RevisionDate") 
119    elem = SubElement(date,"Year") 
120    elem.text=str(tm[0]) 
121    elem = SubElement(date,"Month") 
122    elem.text=str(tm[1]) 
123    elem = SubElement(date,"Day") 
124    elem.text=str(tm[2]) 
125    note = SubElement(revision,"RevisionNote") 
126    note.text = "Created" 
127   
128    return ElementTree(head) 
129   
130   
131  if __name__=='__main__': 
132    import sys,cPickle 
133    from cStringIO import StringIO 
134    pkg = cPickle.load(open(sys.argv[1],'rb')) 
135    perf = (.80,.95,.70,[[4,1],[1,4]]) 
136    tree = PackageToXml(pkg,dataPerformance=[('training data performance',perf)]) 
137    io = StringIO() 
138    tree.write(io) 
139    txt = io.getvalue() 
140    header = """<?xml version="1.0"?> 
141  <!DOCTYPE RDModelInfo PUBLIC "-//RD//DTD RDModelInfo //EN" "RD_Model.dtd"> 
142  """ 
143    print header 
144    print txt.replace('><','>\n<') 
145