pyscinloopsr.sr

  1"""
  2Module to fit PySR and perform analysis on the results
  3"""
  4
  5from pysr import PySRRegressor
  6import joblib
  7from datetime import datetime
  8import pandas as pd
  9import numpy as np
 10from scipy import stats
 11import seaborn as snsi
 12import matplotlib.pyplot as plt
 13from sklearn.inspection import PartialDependenceDisplay
 14from matplotlib.pylab import rcParams
 15rcParams["figure.figsize"] = 15,6
 16plt.figure(figsize=(15, 10))
 17
 18
 19class pysr:
 20    """Class for handling PySR interactions"""
 21
 22    def __init__(self, modelObjtRef):
 23        """
 24        Initiate the object with your PySRRegressor defined object\n
 25        Example:\n
 26            from pyscinloopsr import sr\n
 27            pysrHandlerObject = sr.pysr(PySRRegressorObject)
 28        """
 29        self.model = modelObjtRef
 30        """Internal model reference"""
 31        self.dfxn = None
 32        """Internal predictors Dataframe name"""
 33        self.dfyn = None
 34        """Internal target Dataframe name"""
 35
 36
 37    def runPySR(self, dfx, dfy, featuresDrop=None):
 38        """
 39        Method to fit the PySR model\n
 40        Args: predictors DataFrame, target DataFrame, optional list of features names to drop\n
 41        Example:\n
 42            pysrHandlerObject.runPySR(X, y)\n
 43        Example:\n
 44            pysrHandlerObject.runPySR(X, y, ["featureNameOne", "featureNameTwo"])    
 45        """       
 46        self.dfxn = dfx.copy()
 47        self.dfyn = dfy.copy()
 48        if featuresDrop is not None:
 49            for i in featuresDrop:
 50                print("\ndropping: ", i)
 51                self.dfxn = self.dfxn.drop(i, axis=1)
 52
 53
 54        self.model.fit(self.dfxn, self.dfyn.values.ravel())
 55    
 56
 57    def exportModel(self, directory):
 58        """
 59        Method to export the fitted PySR model\n
 60        Args: directory to place the file\n
 61        File name is set automatically as YYYYMMDDHHMISS.joblib (e.g. 20240427124411.joblib)\n
 62        Example:\n
 63            pysrHandlerObject.exportModel("/content/")
 64        """        
 65        now = datetime.now()
 66        time = now.strftime("%Y%m%d%H%M%S")
 67        joblib.dump(self.model, directory + time + ".joblib")
 68 
 69    def importModel(self, fullPath):
 70        """
 71        Method to import a PySR model previously exported\n
 72        Args: full directory with file name to recover the file\n
 73        File name is set automatically as YYYYMMDDHHMISS.joblib (e.g. 20240427124411.joblib)\n
 74        Example:\n
 75            pysrHandlerObject.importModel("/content/20240427124411.joblib")
 76        """           
 77        self.model = joblib.load(fullPath)
 78
 79    def setModelToAccuracy(self):
 80        """
 81        Method to change PySR mode to accuracy\n
 82        Example:\n
 83            pysrHandlerObject.setModelToBest()
 84        """
 85        self.model.model_selection = "accuracy"
 86
 87    def setModelToBest(self):
 88        """
 89        Method to change PySR mode to best\n
 90        Example:\n
 91            pysrHandlerObject.setModelToBest()
 92        """        
 93        self.model.model_selection = "best"
 94
 95    def printEquationFormats(self):
 96        """
 97        Method to print the generated equation in three formats (equation, latex, sympy)\n
 98        Example:\n
 99            pysrHandlerObject.printEquationFormats()
100        """          
101        print("equation:\n", self.model.get_best().equation, "\n")
102        print("latex:\n", self.model.latex(), "\n") 
103        print("sympy:\n", self.model.sympy(), "\n")   
104
105    def score(self):
106        """
107        Method to print the model score\n
108        Example:\n
109            pysrHandlerObject.score()
110        """            
111        print(self.model.score(self.dfxn, self.dfyn))
112
113    def compare(self):
114        """
115        Method to scatterplot the True target value against model Prediction\n
116        Example:\n
117            pysrHandlerObject.compare()
118        """           
119        plt.figure(figsize=(5, 5))
120        plt.scatter(self.dfyn, self.model.predict(self.dfxn))
121        plt.xlabel('Truth')
122        plt.ylabel('Prediction')
123        plt.show()
124
125    def compareKde(self):
126        """
127        Method to plot KDE of True target value against model Prediction\n
128        Example:\n
129            pysrHandlerObject.compareKde()
130        """         
131        rcParams["figure.figsize"] = 8,8
132        plt.figure(figsize=(8, 8))
133        realXpredicted = self.dfyn.copy()
134        realXpredicted.columns = ["target"]
135        realXpredicted["predicted"] = self.model.predict(self.dfxn)
136        minAux = realXpredicted.target.min()
137        min = minAux
138        minAux = realXpredicted.predicted.min()
139        if minAux<min:
140            min=minAux
141        maxAux = realXpredicted.target.max()
142        max = maxAux
143        maxAux = realXpredicted.predicted.max()
144        if maxAux>max:
145            max=maxAux
146        xAxis = np.linspace(min, max, num=200)
147        ax = realXpredicted.plot.kde(ind=xAxis)
148        plt.xlabel("Values")
149        plt.ylabel("Density")
150        plt.grid(False)
151        plt.show()
152
153
154    def predictCase(self, example):
155        """
156        Method to predict an instance with user defined values for probing the mode\n
157        Example:\n
158            pysrHandlerObject.predictCase([1.67, 0.9, -0.77, -1.7, -0.6, 0.03, 1.04, -0.91,	1.42, -0.91])
159        """          
160        print(self.model.predict(np.array(example).reshape(1, -1)))     
161
162
163    def pdp(self, featureList=None):
164        """
165        Method to print PDP\n
166        Args: Optional list of features names to plot\n
167        Example (all features):\n
168            pysrHandlerObject.pdp()\n
169        Example (specified features):\n
170            pysrHandlerObject.pdp(["featureNameOne", "featureNameTwo", "featureNameThree"])    
171        """       
172        if featureList is None:
173            allFeatures = list(range(self.dfxn.shape[1]))
174        else:
175            allFeatures = featureList
176        fig, ax = plt.subplots(figsize=(10, 20))
177        ax.set_title("Partial Dependence Plots")
178        PartialDependenceDisplay.from_estimator(
179            estimator=self.model,
180            X=self.dfxn,
181            features=(allFeatures),
182            random_state=42,
183            ax=ax
184        )
185        plt.show()