alu/tra-analysis
1
0
Fork 0
mirror of https://github.com/titanscouting/tra-analysis.git synced 2024-12-26 17:49:09 +00:00
Code Issues Packages Projects Releases Wiki Activity
da74e45c9f
tra-analysis/analysis-master/tra_analysis/Analysis.py
Arthur Lu 30a2334d54 fixed import error in Analysis, 
updated unit tests
2021-11-18 09:51:39 +00:00

704 lines
22 KiB
Python
Raw Blame History

# Titan Robotics Team 2022: Analysis Module
# Written by Arthur Lu
# Notes:
# this should be imported as a python module using 'from tra_analysis import Analysis'
# this should be included in the local directory or environment variable
# this module has been optimized for multhreaded computing
# current benchmark of optimization: 1.33 times faster
# setup:
__version__ = "3.0.6"
# changelog should be viewed using print(analysis.__changelog__)
__changelog__ = """changelog:
3.0.6:
- added docstrings
3.0.5:
- removed extra submodule imports
- fixed/optimized header
3.0.4:
- removed -_obj imports
3.0.3:
- fixed spelling of deprecate
3.0.2:
- fixed __all__
3.0.1:
- removed numba dependency and calls
3.0.0:
- exported several submodules to their own files while preserving backwards compatibility:
- Array
- ClassificationMetric
- CorrelationTest
- KNN
- NaiveBayes
- RandomForest
- RegressionMetric
- Sort
- StatisticalTest
- SVM
- note: above listed submodules will not be supported in the future
- future changes to all submodules will be held in their respective changelogs
- future changes altering the parent package will be held in the __changelog__ of the parent package (in __init__.py)
- changed reference to module name to Analysis
2.3.1:
- fixed bugs in Array class
2.3.0:
- overhauled Array class
2.2.3:
- fixed spelling of RandomForest
- made n_neighbors required for KNN
- made n_classifiers required for SVM
2.2.2:
- fixed 2.2.1 changelog entry
- changed regression to return dictionary
2.2.1:
- changed all references to parent package analysis to tra_analysis
2.2.0:
- added Sort class
- added several array sorting functions to Sort class including:
- quick sort
- merge sort
- intro(spective) sort
- heap sort
- insertion sort
- tim sort
- selection sort
- bubble sort
- cycle sort
- cocktail sort
- tested all sorting algorithms with both lists and numpy arrays
- deprecated sort function from Array class
- added warnings as an import
2.1.4:
- added sort and search functions to Array class
2.1.3:
- changed output of basic_stats and histo_analysis to libraries
- fixed __all__
2.1.2:
- renamed ArrayTest class to Array
2.1.1:
- added add, mul, neg, and inv functions to ArrayTest class
- added normalize function to ArrayTest class
- added dot and cross functions to ArrayTest class
2.1.0:
- added ArrayTest class
- added elementwise mean, median, standard deviation, variance, min, max functions to ArrayTest class
- added elementwise_stats to ArrayTest which encapsulates elementwise statistics
- appended to __all__ to reflect changes
2.0.6:
- renamed func functions in regression to lin, log, exp, and sig
2.0.5:
- moved random_forrest_regressor and random_forrest_classifier to RandomForrest class
- renamed Metrics to Metric
- renamed RegressionMetrics to RegressionMetric
- renamed ClassificationMetrics to ClassificationMetric
- renamed CorrelationTests to CorrelationTest
- renamed StatisticalTests to StatisticalTest
- reflected rafactoring to all mentions of above classes/functions
2.0.4:
- fixed __all__ to reflected the correct functions and classes
- fixed CorrelationTests and StatisticalTests class functions to require self invocation
- added missing math import
- fixed KNN class functions to require self invocation
- fixed Metrics class functions to require self invocation
- various spelling fixes in CorrelationTests and StatisticalTests
2.0.3:
- bug fixes with CorrelationTests and StatisticalTests
- moved glicko2 and trueskill to the metrics subpackage
- moved elo to a new metrics subpackage
2.0.2:
- fixed docs
2.0.1:
- fixed docs
2.0.0:
- cleaned up wild card imports with scipy and sklearn
- added CorrelationTests class
- added StatisticalTests class
- added several correlation tests to CorrelationTests
- added several statistical tests to StatisticalTests
1.13.9:
- moved elo, glicko2, trueskill functions under class Metrics
1.13.8:
- moved Glicko2 to a seperate package
1.13.7:
- fixed bug with trueskill
1.13.6:
- cleaned up imports
1.13.5:
- cleaned up package
1.13.4:
- small fixes to regression to improve performance
1.13.3:
- filtered nans from regression
1.13.2:
- removed torch requirement, and moved Regression back to regression.py
1.13.1:
- bug fix with linear regression not returning a proper value
- cleaned up regression
- fixed bug with polynomial regressions
1.13.0:
- fixed all regressions to now properly work
1.12.6:
- fixed bg with a division by zero in histo_analysis
1.12.5:
- fixed numba issues by removing numba from elo, glicko2 and trueskill
1.12.4:
- renamed gliko to glicko
1.12.3:
- removed deprecated code
1.12.2:
- removed team first time trueskill instantiation in favor of integration in superscript.py
1.12.1:
- improved readibility of regression outputs by stripping tensor data
- used map with lambda to acheive the improved readibility
- lost numba jit support with regression, and generated_jit hangs at execution
- TODO: reimplement correct numba integration in regression
1.12.0:
- temporarily fixed polynomial regressions by using sklearn's PolynomialFeatures
1.11.010:
- alphabeticaly ordered import lists
1.11.9:
- bug fixes
1.11.8:
- bug fixes
1.11.7:
- bug fixes
1.11.6:
- tested min and max
- bug fixes
1.11.5:
- added min and max in basic_stats
1.11.4:
- bug fixes
1.11.3:
- bug fixes
1.11.2:
- consolidated metrics
- fixed __all__
1.11.1:
- added test/train split to RandomForestClassifier and RandomForestRegressor
1.11.0:
- added RandomForestClassifier and RandomForestRegressor
- note: untested
1.10.0:
- added numba.jit to remaining functions
1.9.2:
- kernelized PCA and KNN
1.9.1:
- fixed bugs with SVM and NaiveBayes
1.9.0:
- added SVM class, subclasses, and functions
- note: untested
1.8.0:
- added NaiveBayes classification engine
- note: untested
1.7.0:
- added knn()
- added confusion matrix to decisiontree()
1.6.2:
- changed layout of __changelog to be vscode friendly
1.6.1:
- added additional hyperparameters to decisiontree()
1.6.0:
- fixed __version__
- fixed __all__ order
- added decisiontree()
1.5.3:
- added pca
1.5.2:
- reduced import list
- added kmeans clustering engine
1.5.1:
- simplified regression by using .to(device)
1.5.0:
- added polynomial regression to regression(); untested
1.4.0:
- added trueskill()
1.3.2:
- renamed regression class to Regression, regression_engine() to regression gliko2_engine class to Gliko2
1.3.1:
- changed glicko2() to return tuple instead of array
1.3.0:
- added glicko2_engine class and glicko()
- verified glicko2() accuracy
1.2.3:
- fixed elo()
1.2.2:
- added elo()
- elo() has bugs to be fixed
1.2.1:
- readded regrression import
1.2.0:
- integrated regression.py as regression class
- removed regression import
- fixed metadata for regression class
- fixed metadata for analysis class
1.1.1:
- regression_engine() bug fixes, now actaully regresses
1.1.0:
- added regression_engine()
- added all regressions except polynomial
1.0.7:
- updated _init_device()
1.0.6:
- removed useless try statements
1.0.5:
- removed impossible outcomes
1.0.4:
- added performance metrics (r^2, mse, rms)
1.0.3:
- resolved nopython mode for mean, median, stdev, variance
1.0.2:
- snapped (removed) majority of uneeded imports
- forced object mode (bad) on all jit
- TODO: stop numba complaining about not being able to compile in nopython mode
1.0.1:
- removed from sklearn import * to resolve uneeded wildcard imports
1.0.0:
- removed c_entities,nc_entities,obstacles,objectives from __all__
- applied numba.jit to all functions
- deprecated and removed stdev_z_split
- cleaned up histo_analysis to include numpy and numba.jit optimizations
- deprecated and removed all regression functions in favor of future pytorch optimizer
- deprecated and removed all nonessential functions (basic_analysis, benchmark, strip_data)
- optimized z_normalize using sklearn.preprocessing.normalize
- TODO: implement kernel/function based pytorch regression optimizer
0.9.0:
- refactored
- numpyed everything
- removed stats in favor of numpy functions
0.8.5:
- minor fixes
0.8.4:
- removed a few unused dependencies
0.8.3:
- added p_value function
0.8.2:
- updated __all__ correctly to contain changes made in v 0.8.0 and v 0.8.1
0.8.1:
- refactors
- bugfixes
0.8.0:
- deprecated histo_analysis_old
- deprecated debug
- altered basic_analysis to take array data instead of filepath
- refactor
- optimization
0.7.2:
- bug fixes
0.7.1:
- bug fixes
0.7.0:
- added tanh_regression (logistical regression)
- bug fixes
0.6.5:
- added z_normalize function to normalize dataset
- bug fixes
0.6.4:
- bug fixes
0.6.3:
- bug fixes
0.6.2:
- bug fixes
0.6.1:
- corrected __all__ to contain all of the functions
0.6.0:
- added calc_overfit, which calculates two measures of overfit, error and performance
- added calculating overfit to optimize_regression
0.5.0:
- added optimize_regression function, which is a sample function to find the optimal regressions
- optimize_regression function filters out some overfit funtions (functions with r^2 = 1)
- planned addition: overfit detection in the optimize_regression function
0.4.2:
- added __changelog__
- updated debug function with log and exponential regressions
0.4.1:
- added log regressions
- added exponential regressions
- added log_regression and exp_regression to __all__
0.3.8:
- added debug function to further consolidate functions
0.3.7:
- added builtin benchmark function
- added builtin random (linear) data generation function
- added device initialization (_init_device)
0.3.6:
- reorganized the imports list to be in alphabetical order
- added search and regurgitate functions to c_entities, nc_entities, obstacles, objectives
0.3.5:
- major bug fixes
- updated historical analysis
- deprecated old historical analysis
0.3.4:
- added __version__, __author__, __all__
- added polynomial regression
- added root mean squared function
- added r squared function
0.3.3:
- bug fixes
- added c_entities
0.3.2:
- bug fixes
- added nc_entities, obstacles, objectives
- consolidated statistics.py to analysis.py
0.3.1:
- compiled 1d, column, and row basic stats into basic stats function
0.3.0:
- added historical analysis function
0.2.x:
- added z score test
0.1.x:
- major bug fixes
0.0.x:
- added loading csv
- added 1d, column, row basic stats
"""
__author__ = (
"Arthur Lu <learthurgo@gmail.com>",
)
__all__ = [
'load_csv',
'basic_stats',
'z_score',
'z_normalize',
'histo_analysis',
'regression',
'Metric',
'pca',
'decisiontree',
# all statistics functions left out due to integration in other functions
]
# now back to your regularly scheduled programming:
# imports (now in alphabetical order! v 0.3.006):
import csv
from tra_analysis.metrics import elo as Elo
from tra_analysis.metrics import glicko2 as Glicko2
import numpy as np
import scipy
import sklearn, sklearn.cluster, sklearn.pipeline
from tra_analysis.metrics import trueskill as Trueskill
# import submodules
from .ClassificationMetric import ClassificationMetric
class error(ValueError):
pass
def load_csv(filepath):
"""
Loads csv file into 2D numpy array. Does not check csv file validity.
parameters:
filepath: String path to the csv file
return:
2D numpy array of values stored in csv file
"""
with open(filepath, newline='') as csvfile:
file_array = np.array(list(csv.reader(csvfile)))
csvfile.close()
return file_array
def basic_stats(data):
"""
Calculates mean, median, standard deviation, variance, minimum, maximum of a simple set of elements.
parameters:
data: List representing set of unordered elements
return:
Dictionary with (mean, median, standard-deviation, variance, minimum, maximum) as keys and corresponding values
"""
data_t = np.array(data).astype(float)
_mean = mean(data_t)
_median = median(data_t)
_stdev = stdev(data_t)
_variance = variance(data_t)
_min = npmin(data_t)
_max = npmax(data_t)
return {"mean": _mean, "median": _median, "standard-deviation": _stdev, "variance": _variance, "minimum": _min, "maximum": _max}
def z_score(point, mean, stdev):
"""
Calculates z score of a specific point given mean and standard deviation of data.
parameters:
point: Real value corresponding to a single point of data
mean: Real value corresponding to the mean of the dataset
stdev: Real value corresponding to the standard deviation of the dataset
return:
Real value that is the point's z score
"""
score = (point - mean) / stdev
return score
def z_normalize(array, *args):
"""
Applies sklearn.normalize(array, axis = args) on any arraylike parseable by numpy.
parameters:
array: array like structure of reals aka nested indexables
*args: arguments relating to axis normalized against
return:
numpy array of normalized values from ArrayLike input
"""
array = np.array(array)
for arg in args:
array = sklearn.preprocessing.normalize(array, axis = arg)
return array
def histo_analysis(hist_data):
"""
Calculates the mean and standard deviation of derivatives of (x,y) points. Requires at least 2 points to compute.
parameters:
hist_data: list of real coordinate point data (x, y)
return:
Dictionary with (mean, deviation) as keys to corresponding values
"""
if len(hist_data[0]) > 2:
hist_data = np.array(hist_data)
derivative = np.array(len(hist_data) - 1, dtype = float)
t = np.diff(hist_data)
derivative = t[1] / t[0]
np.sort(derivative)
return {"mean": basic_stats(derivative)["mean"], "deviation": basic_stats(derivative)["standard-deviation"]}
else:
return None
def regression(inputs, outputs, args): # inputs, outputs expects N-D array
"""
Applies specified regression kernels onto input, output data pairs.
parameters:
inputs: List of Reals representing independent variable values of each point
outputs: List of Reals representing dependent variable values of each point
args: List of Strings from values (lin, log, exp, ply, sig)
return:
Dictionary with keys (lin, log, exp, ply, sig) as keys to correspondiong regression models
"""
X = np.array(inputs)
y = np.array(outputs)
regressions = {}
if 'lin' in args: # formula: ax + b
try:
def lin(x, a, b):
return a * x + b
popt, pcov = scipy.optimize.curve_fit(lin, X, y)
coeffs = popt.flatten().tolist()
regressions["lin"] = (str(coeffs[0]) + "*x+" + str(coeffs[1]))
except Exception as e:
pass
if 'log' in args: # formula: a log (b(x + c)) + d
try:
def log(x, a, b, c, d):
return a * np.log(b*(x + c)) + d
popt, pcov = scipy.optimize.curve_fit(log, X, y)
coeffs = popt.flatten().tolist()
regressions["log"] = (str(coeffs[0]) + "*log(" + str(coeffs[1]) + "*(x+" + str(coeffs[2]) + "))+" + str(coeffs[3]))
except Exception as e:
pass
if 'exp' in args: # formula: a e ^ (b(x + c)) + d
try:
def exp(x, a, b, c, d):
return a * np.exp(b*(x + c)) + d
popt, pcov = scipy.optimize.curve_fit(exp, X, y)
coeffs = popt.flatten().tolist()
regressions["exp"] = (str(coeffs[0]) + "*e^(" + str(coeffs[1]) + "*(x+" + str(coeffs[2]) + "))+" + str(coeffs[3]))
except Exception as e:
pass
if 'ply' in args: # formula: a + bx^1 + cx^2 + dx^3 + ...
inputs = np.array([inputs])
outputs = np.array([outputs])
plys = {}
limit = len(outputs[0])
for i in range(2, limit):
model = sklearn.preprocessing.PolynomialFeatures(degree = i)
model = sklearn.pipeline.make_pipeline(model, sklearn.linear_model.LinearRegression())
model = model.fit(np.rot90(inputs), np.rot90(outputs))
params = model.steps[1][1].intercept_.tolist()
params = np.append(params, model.steps[1][1].coef_[0].tolist()[1::])
params = params.flatten().tolist()
temp = ""
counter = 0
for param in params:
temp += "(" + str(param) + "*x^" + str(counter) + ")"
counter += 1
plys["x^" + str(i)] = (temp)
regressions["ply"] = (plys)
if 'sig' in args: # formula: a tanh (b(x + c)) + d
try:
def sig(x, a, b, c, d):
return a * np.tanh(b*(x + c)) + d
popt, pcov = scipy.optimize.curve_fit(sig, X, y)
coeffs = popt.flatten().tolist()
regressions["sig"] = (str(coeffs[0]) + "*tanh(" + str(coeffs[1]) + "*(x+" + str(coeffs[2]) + "))+" + str(coeffs[3]))
except Exception as e:
pass
return regressions
class Metric:
"""
The metric class wraps the metrics models. Call without instantiation as Metric.<method>(...)
"""
def elo(self, starting_score, opposing_score, observed, N, K):
"""
Calculates an elo adjusted ELO score given a player's current score, opponent's score, and outcome of match.
reference: https://en.wikipedia.org/wiki/Elo_rating_system
parameters:
starting_score: Real value representing player's ELO score before a match
opposing_score: Real value representing opponent's score before the match
observed: Array of Real values representing multiple sequential match outcomes against the same opponent. 1 for match win, 0.5 for tie, 0 for loss.
N: Real value representing the normal or mean score expected (usually 1200)
K: R eal value representing a system constant, determines how quickly players will change scores (usually 24)
return:
Real value representing the player's new ELO score
"""
return Elo.calculate(starting_score, opposing_score, observed, N, K)
def glicko2(self, starting_score, starting_rd, starting_vol, opposing_score, opposing_rd, observations):
"""
Calculates an adjusted Glicko-2 score given a player's current score, multiple opponent's score, and outcome of several matches.
reference: http://www.glicko.net/glicko/glicko2.pdf
parameters:
starting_score: Real value representing the player's Glicko-2 score
starting_rd: Real value representing the player's RD
starting_vol: Real value representing the player's volatility
opposing_score: List of Real values representing multiple opponent's Glicko-2 scores
opposing_rd: List of Real values representing multiple opponent's RD
opposing_vol: List of Real values representing multiple opponent's volatility
observations: List of Real values representing the outcome of several matches, where each match's opponent corresponds with the opposing_score, opposing_rd, opposing_vol values of the same indesx. Outcomes can be a score, presuming greater score is better.
return:
Tuple of 3 Real values representing the player's new score, rd, and vol
"""
player = Glicko2.Glicko2(rating = starting_score, rd = starting_rd, vol = starting_vol)
player.update_player([x for x in opposing_score], [x for x in opposing_rd], observations)
return (player.rating, player.rd, player.vol)
def trueskill(self, teams_data, observations): # teams_data is array of array of tuples ie. [[(mu, sigma), (mu, sigma), (mu, sigma)], [(mu, sigma), (mu, sigma), (mu, sigma)]]
"""
Calculates the score changes for multiple teams playing in a single match accoding to the trueskill algorithm.
reference: https://trueskill.org/
parameters:
teams_data: List of List of Tuples of 2 Real values representing multiple player ratings. List of teams, which is a List of players. Each player rating is a Tuple of 2 Real values (mu, sigma).
observations: List of Real values representing the match outcome. Each value in the List is the score corresponding to the team at the same index in teams_data.
return:
List of List of Tuples of 2 Real values representing new player ratings. Same structure as teams_data.
"""
team_ratings = []
for team in teams_data:
team_temp = ()
for player in team:
player = Trueskill.Rating(player[0], player[1])
team_temp = team_temp + (player,)
team_ratings.append(team_temp)
return Trueskill.rate(team_ratings, ranks=observations)
def mean(data):
return np.mean(data)
def median(data):
return np.median(data)
def stdev(data):
return np.std(data)
def variance(data):
return np.var(data)
def npmin(data):
return np.amin(data)
def npmax(data):
return np.amax(data)
def pca(data, n_components = None, copy = True, whiten = False, svd_solver = "auto", tol = 0.0, iterated_power = "auto", random_state = None):
"""
Performs a principle component analysis on the input data.
reference: https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html
parameters:
data: Arraylike of Reals representing the set of data to perform PCA on
* : refer to reference for usage, parameters follow same usage
return:
Arraylike of Reals representing the set of data that has had PCA performed. The dimensionality of the Arraylike may be smaller or equal.
"""
kernel = sklearn.decomposition.PCA(n_components = n_components, copy = copy, whiten = whiten, svd_solver = svd_solver, tol = tol, iterated_power = iterated_power, random_state = random_state)
return kernel.fit_transform(data)
def decisiontree(data, labels, test_size = 0.3, criterion = "gini", splitter = "default", max_depth = None): #expects *2d data and 1d labels
"""
Generates a decision tree classifier fitted to the given data.
reference: https://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html
parameters:
data: List of values representing each data point of multiple axes
labels: List of values represeing the labels corresponding to the same index at data
* : refer to reference for usage, parameters follow same usage
return:
DecisionTreeClassifier model and corresponding classification accuracy metrics
"""
data_train, data_test, labels_train, labels_test = sklearn.model_selection.train_test_split(data, labels, test_size=test_size, random_state=1)
model = sklearn.tree.DecisionTreeClassifier(criterion = criterion, splitter = splitter, max_depth = max_depth)
model = model.fit(data_train,labels_train)
predictions = model.predict(data_test)
metrics = ClassificationMetric(predictions, labels_test)
return model, metrics
Reference in New Issue View Git Blame Copy Permalink