mirror of
https://github.com/titanscouting/tra-analysis.git
synced 2024-11-10 06:54:44 +00:00
analysis-better.py v 1.0.9.000
changelog: - refactored - numpyed everything - removed stats in favor of numpy functions
This commit is contained in:
parent
e914d32b37
commit
b689dada3d
@ -14,6 +14,7 @@ __changelog__ = """changelog:
|
||||
1.0.9.000:
|
||||
- refactored
|
||||
- numpyed everything
|
||||
- removed stats in favor of numpy functions
|
||||
1.0.8.005:
|
||||
- minor fixes
|
||||
1.0.8.004:
|
||||
@ -160,7 +161,6 @@ import torch
|
||||
class error(ValueError):
|
||||
pass
|
||||
|
||||
|
||||
def _init_device(setting, arg): # initiates computation device for ANNs
|
||||
if setting == "cuda":
|
||||
try:
|
||||
@ -177,11 +177,10 @@ def _init_device(setting, arg): # initiates computation device for ANNs
|
||||
|
||||
def load_csv(filepath):
|
||||
with open(filepath, newline='') as csvfile:
|
||||
file_array = list(csv.reader(csvfile))
|
||||
file_array = np.array(list(csv.reader(csvfile)))
|
||||
csvfile.close()
|
||||
return file_array
|
||||
|
||||
|
||||
# data=array, mode = ['1d':1d_basic_stats, 'column':c_basic_stats, 'row':r_basic_stats], arg for mode 1 or mode 2 for column or row
|
||||
def basic_stats(data, method, arg):
|
||||
|
||||
@ -190,10 +189,7 @@ def basic_stats(data, method, arg):
|
||||
|
||||
if method == "1d" or method == 0:
|
||||
|
||||
data_t = []
|
||||
|
||||
for i in range(0, len(data), 1):
|
||||
data_t.append(float(data[i]))
|
||||
data_t = np.array(data).astype(float)
|
||||
|
||||
_mean = mean(data_t)
|
||||
_median = median(data_t)
|
||||
@ -211,7 +207,7 @@ def basic_stats(data, method, arg):
|
||||
_variance = None
|
||||
|
||||
return _mean, _median, _mode, _stdev, _variance
|
||||
|
||||
"""
|
||||
elif method == "column" or method == 1:
|
||||
|
||||
c_data = []
|
||||
@ -239,7 +235,7 @@ def basic_stats(data, method, arg):
|
||||
_variance = None
|
||||
|
||||
return _mean, _median, _mode, _stdev, _variance
|
||||
|
||||
|
||||
elif method == "row" or method == 2:
|
||||
|
||||
r_data = []
|
||||
@ -263,9 +259,10 @@ def basic_stats(data, method, arg):
|
||||
_variance = None
|
||||
|
||||
return _mean, _median, _mode, _stdev, _variance
|
||||
|
||||
|
||||
else:
|
||||
raise error("method error")
|
||||
"""
|
||||
|
||||
|
||||
# returns z score with inputs of point, mean and standard deviation of spread
|
||||
@ -277,8 +274,8 @@ def z_score(point, mean, stdev):
|
||||
# mode is either 'x' or 'y' or 'both' depending on the variable(s) to be normalized
|
||||
def z_normalize(x, y, mode):
|
||||
|
||||
x_norm = []
|
||||
y_norm = []
|
||||
x_norm = np.array().astype(float)
|
||||
y_norm = np.array().astype(float)
|
||||
|
||||
mean = 0
|
||||
stdev = 0
|
||||
@ -320,7 +317,7 @@ def z_normalize(x, y, mode):
|
||||
# returns n-th percentile of spread given mean, standard deviation, lower z-score, and upper z-score
|
||||
def stdev_z_split(mean, stdev, delta, low_bound, high_bound):
|
||||
|
||||
z_split = []
|
||||
z_split = np.array().astype(float)
|
||||
i = low_bound
|
||||
|
||||
while True:
|
||||
@ -715,6 +712,27 @@ def generate_data(filename, x, y, low, high):
|
||||
temp = temp + str(random.uniform(low, high))
|
||||
file.write(temp + "\n")
|
||||
|
||||
def mean(data):
|
||||
|
||||
return np.mean(data)
|
||||
|
||||
def median(data):
|
||||
|
||||
return np.median(data)
|
||||
|
||||
def mode(data):
|
||||
|
||||
return np.argmax(np.bincount(data))
|
||||
|
||||
def stdev(data):
|
||||
|
||||
return np.std(data)
|
||||
|
||||
def variance(data):
|
||||
|
||||
return np.var(data)
|
||||
|
||||
"""
|
||||
|
||||
class StatisticsError(ValueError):
|
||||
pass
|
||||
@ -856,8 +874,6 @@ def _fail_neg(values, errmsg='negative value'):
|
||||
if x < 0:
|
||||
raise StatisticsError(errmsg)
|
||||
yield x
|
||||
|
||||
|
||||
def mean(data):
|
||||
|
||||
if iter(data) is data:
|
||||
@ -927,3 +943,4 @@ def stdev(data, xbar=None):
|
||||
return var.sqrt()
|
||||
except AttributeError:
|
||||
return math.sqrt(var)
|
||||
"""
|
Loading…
Reference in New Issue
Block a user