tsadmetrics 0.1.8__tar.gz → 0.1.9__tar.gz

{tsadmetrics-0.1.8/tsadmetrics.egg-info → tsadmetrics-0.1.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tsadmetrics
-Version: 0.1.8
+Version: 0.1.9
 Summary: Librería para evaluación de detección de anomalías en series temporales
 Home-page: https://github.com/pathsko/TSADmetrics
 Author: Pedro Rafael Velasco Priego

{tsadmetrics-0.1.8 → tsadmetrics-0.1.9}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "tsadmetrics"
-version = "0.1.8"
+version = "0.1.9"
 description = "Librería para evaluación de detección de anomalías en series temporales"
 authors = [
   { name = "Pedro Rafael Velasco Priego", email = "i12veprp@uco.es" }

{tsadmetrics-0.1.8 → tsadmetrics-0.1.9}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r", encoding="utf-8") as fh:
 
 setup(
     name="tsadmetrics",
-    version="0.1.8",
+    version="0.1.9",
     author="Pedro Rafael Velasco Priego",
     author_email="i12veprp@uco.es",
     description="A library for time series anomaly detection metrics and evaluation.",

{tsadmetrics-0.1.8 → tsadmetrics-0.1.9}/tsadmetrics/_tsadeval/affiliation/_integral_interval.py

@@ -125,13 +125,13 @@ def get_pivot_j(I, J):
 def integral_mini_interval(I, J):
     """
     In the specific case where interval I is located outside J,
-    integral of distance from x to J over the interval x \in I.
+    integral of distance from x to J over the interval x in I.
     This is the *integral* i.e. the sum.
     It's not the mean (not divided by the length of I yet)
     
     :param I: a interval (start, stop), or None
     :param J: a non empty interval, with empty intersection with I
-    :return: the integral of distances d(x, J) over x \in I
+    :return: the integral of distances d(x, J) over x in I
     """
     if I is None:
         return(0)
@@ -144,14 +144,14 @@ def integral_mini_interval(I, J):
 def integral_interval_distance(I, J):
     """
     For any non empty intervals I, J, compute the
-    integral of distance from x to J over the interval x \in I.
+    integral of distance from x to J over the interval x in I.
     This is the *integral* i.e. the sum. 
     It's not the mean (not divided by the length of I yet)
     The interval I can intersect J or not
     
     :param I: a interval (start, stop), or None
     :param J: a non empty interval
-    :return: the integral of distances d(x, J) over x \in I
+    :return: the integral of distances d(x, J) over x in I
     """
     # I and J are single intervals (not generic sets)
     # I is a predicted interval in the range of affiliation of J
@@ -178,15 +178,15 @@ def integral_mini_interval_P_CDFmethod__min_piece(I, J, E):
     """
     Helper of `integral_mini_interval_Pprecision_CDFmethod`
     In the specific case where interval I is located outside J,
-    compute the integral $\int_{d_min}^{d_max} \min(m, x) dx$, with:
+    compute the integral $int_{d_min}^{d_max} min(m, x) dx$, with:
     - m the smallest distance from J to E,
-    - d_min the smallest distance d(x, J) from x \in I to J
-    - d_max the largest distance d(x, J) from x \in I to J
+    - d_min the smallest distance d(x, J) from x in I to J
+    - d_max the largest distance d(x, J) from x in I to J
     
     :param I: a single predicted interval, a non empty interval (start, stop)
     :param J: ground truth interval, a non empty interval, with empty intersection with I
     :param E: the affiliation/influence zone for J, represented as a couple (start, stop)
-    :return: the integral $\int_{d_min}^{d_max} \min(m, x) dx$
+    :return: the integral $int_{d_min}^{d_max} min(m, x) dx$
     """
     if interval_intersection(I, J) is not None:
         raise ValueError('I and J should have a void intersection')
@@ -214,13 +214,13 @@ def integral_mini_interval_Pprecision_CDFmethod(I, J, E):
     """
     Integral of the probability of distances over the interval I.
     In the specific case where interval I is located outside J,
-    compute the integral $\int_{x \in I} Fbar(dist(x,J)) dx$.
+    compute the integral $int_{x in I} Fbar(dist(x,J)) dx$.
     This is the *integral* i.e. the sum (not the mean)
     
     :param I: a single predicted interval, a non empty interval (start, stop)
     :param J: ground truth interval, a non empty interval, with empty intersection with I
     :param E: the affiliation/influence zone for J, represented as a couple (start, stop)
-    :return: the integral $\int_{x \in I} Fbar(dist(x,J)) dx$
+    :return: the integral $int_{x in I} Fbar(dist(x,J)) dx$
     """
     integral_min_piece = integral_mini_interval_P_CDFmethod__min_piece(I, J, E)
   
@@ -244,13 +244,13 @@ def integral_mini_interval_Pprecision_CDFmethod(I, J, E):
 def integral_interval_probaCDF_precision(I, J, E):
     """
     Integral of the probability of distances over the interval I.
-    Compute the integral $\int_{x \in I} Fbar(dist(x,J)) dx$.
+    Compute the integral $int_{x in I} Fbar(dist(x,J)) dx$.
     This is the *integral* i.e. the sum (not the mean)
     
     :param I: a single (non empty) predicted interval in the zone of affiliation of J
     :param J: ground truth interval
     :param E: affiliation/influence zone for J
-    :return: the integral $\int_{x \in I} Fbar(dist(x,J)) dx$
+    :return: the integral $int_{x in I} Fbar(dist(x,J)) dx$
     """
     # I and J are single intervals (not generic sets)
     def f(I_cut):
@@ -307,13 +307,13 @@ def integral_mini_interval_Precall_CDFmethod(I, J, E):
     """
     Integral of the probability of distances over the interval J.
     In the specific case where interval J is located outside I,
-    compute the integral $\int_{y \in J} Fbar_y(dist(y,I)) dy$.
+    compute the integral $int_{y in J} Fbar_y(dist(y,I)) dy$.
     This is the *integral* i.e. the sum (not the mean)
     
     :param I: a single (non empty) predicted interval
     :param J: ground truth (non empty) interval, with empty intersection with I
     :param E: the affiliation/influence zone for J, represented as a couple (start, stop)
-    :return: the integral $\int_{y \in J} Fbar_y(dist(y,I)) dy$
+    :return: the integral $int_{y in J} Fbar_y(dist(y,I)) dy$
     """
     # The interval J should be located outside I 
     # (so it's either the left piece or the right piece w.r.t I)
@@ -377,7 +377,7 @@ def integral_mini_interval_Precall_CDFmethod(I, J, E):
     # j_bb_min       j_bb_max j_ba_min       j_ba_max j_ab_min      j_ab_max j_aa_min      j_aa_max
     # (with `b` for before and `a` for after in the previous variable names)
     
-    #                                          vs e_mean  m = min(t-e_min, e_max-t)  d=|i_pivot-t|   min(d,m)                            \int min(d,m)dt   \int d dt        \int_(min(d,m)+d)dt                                    \int_{t \in J}(min(d,m)+d)dt
+    #                                          vs e_mean  m = min(t-e_min, e_max-t)  d=|i_pivot-t|   min(d,m)                            int min(d,m)dt   int d dt        int_(min(d,m)+d)dt                                    int_{t in J}(min(d,m)+d)dt
     # Case J_before_closeE & i_pivot after J   before     t-e_min                    i_pivot-t       min(i_pivot-t,t-e_min) = t-e_min    t^2/2-e_min*t     i_pivot*t-t^2/2  t^2/2-e_min*t+i_pivot*t-t^2/2 = (i_pivot-e_min)*t      (i_pivot-e_min)*tB - (i_pivot-e_min)*tA = (i_pivot-e_min)*(tB-tA)
     # Case J_before_closeI & i_pivot after J   before     t-e_min                    i_pivot-t       min(i_pivot-t,t-e_min) = i_pivot-t  i_pivot*t-t^2/2   i_pivot*t-t^2/2  i_pivot*t-t^2/2+i_pivot*t-t^2/2 = 2*i_pivot*t-t^2      2*i_pivot*tB-tB^2 - 2*i_pivot*tA + tA^2 = 2*i_pivot*(tB-tA) - (tB^2 - tA^2)
     # Case J_after_closeI & i_pivot after J    after      e_max-t                    i_pivot-t       min(i_pivot-t,e_max-t) = i_pivot-t  i_pivot*t-t^2/2   i_pivot*t-t^2/2  i_pivot*t-t^2/2+i_pivot*t-t^2/2 = 2*i_pivot*t-t^2      2*i_pivot*tB-tB^2 - 2*i_pivot*tA + tA^2 = 2*i_pivot*(tB-tA) - (tB^2 - tA^2)
@@ -406,12 +406,12 @@ def integral_mini_interval_Precall_CDFmethod(I, J, E):
     out_integral_min_dm_plus_d = _sum_wo_nan(out_parts) # integral on all J, i.e. sum of the disjoint parts
 
     # We have for each point t of J:
-    # \bar{F}_{t, recall}(d) = 1 - (1/|E|) * (min(d,m) + d)
+    # bar{F}_{t, recall}(d) = 1 - (1/|E|) * (min(d,m) + d)
     # Since t is a single-point here, and we are in the case where i_pivot is inside E.
     # The integral is then given by:
-    # C = \int_{t \in J} \bar{F}_{t, recall}(D(t)) dt
-    #   = \int_{t \in J} 1 - (1/|E|) * (min(d,m) + d) dt
-    #   = |J| - (1/|E|) * [\int_{t \in J} (min(d,m) + d) dt]
+    # C = int_{t in J} bar{F}_{t, recall}(D(t)) dt
+    #   = int_{t in J} 1 - (1/|E|) * (min(d,m) + d) dt
+    #   = |J| - (1/|E|) * [int_{t in J} (min(d,m) + d) dt]
     #   = |J| - (1/|E|) * out_integral_min_dm_plus_d    
     DeltaJ = max(J) - min(J)
     DeltaE = max(E) - min(E)
@@ -422,17 +422,17 @@ def integral_mini_interval_Precall_CDFmethod(I, J, E):
 def integral_interval_probaCDF_recall(I, J, E):
     """
     Integral of the probability of distances over the interval J.
-    Compute the integral $\int_{y \in J} Fbar_y(dist(y,I)) dy$.
+    Compute the integral $int_{y in J} Fbar_y(dist(y,I)) dy$.
     This is the *integral* i.e. the sum (not the mean)
 
     :param I: a single (non empty) predicted interval
     :param J: ground truth (non empty) interval
     :param E: the affiliation/influence zone for J
-    :return: the integral $\int_{y \in J} Fbar_y(dist(y,I)) dy$
+    :return: the integral $int_{y in J} Fbar_y(dist(y,I)) dy$
     """
     # I and J are single intervals (not generic sets)
     # E is the outside affiliation interval of J (even for recall!)
-    # (in particular J \subset E)
+    # (in particular J subset E)
     #
     # J is the portion of the ground truth affiliated to I
     # I is a predicted interval (can be outside E possibly since it's recall)

{tsadmetrics-0.1.8 → tsadmetrics-0.1.9}/tsadmetrics/_tsadeval/nabscore.py

@@ -176,6 +176,8 @@ class Sweeper(object):
           unweightedScore = -1.0
         else:
           numerator = abs(prevWindowRightIndex - i)
+          if prevWindowWidth==1:
+            prevWindowWidth+=1
           denominator = float(prevWindowWidth - 1)
           positionPastWindow = numerator / denominator
           unweightedScore = scaledSigmoid(positionPastWindow)

{tsadmetrics-0.1.8 → tsadmetrics-0.1.9/tsadmetrics.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tsadmetrics
-Version: 0.1.8
+Version: 0.1.9
 Summary: Librería para evaluación de detección de anomalías en series temporales
 Home-page: https://github.com/pathsko/TSADmetrics
 Author: Pedro Rafael Velasco Priego