Use own filter outliers method

nise-circleguard/src/main.py

@@ -21,6 +21,26 @@ cg = Circleguard(os.getenv("OSU_API_KEY"), db_path="./dbs/db.db", slider_dir="./
 
 app = Flask(__name__)
 
+def my_filter_outliers(arr, bias=1.5):
+    """
+    Returns ``arr`` with outliers removed.
+
+    Parameters
+    ----------
+    arr: list
+        List of numbers to filter outliers from.
+    bias: int
+        Points in ``arr`` which are more than ``IQR * bias`` away from the first
+        or third quartile of ``arr`` will be removed.
+    """
+    if not arr or len(arr) <= 0:
+        return arr
+    q3, q1 = np.percentile(arr, [75, 25])
+    iqr = q3 - q1
+    lower_limit = q1 - (bias * iqr)
+    upper_limit = q3 + (bias * iqr)
+    arr_without_outliers = [x for x in arr if lower_limit < x < upper_limit]; return arr if not arr_without_outliers else arr_without_outliers
+
 
 @dataclass
 class ReplayRequest:
@@ -179,15 +199,15 @@ def process_replay():
 
             keypresses_times=kp,
             keypresses_median=np.median(kp),
-            keypresses_median_adjusted=np.median(filter_outliers(kp)),
+            keypresses_median_adjusted=np.median(my_filter_outliers(kp)),
             keypresses_standard_deviation=np.std(kp, ddof=1),
-            keypresses_standard_deviation_adjusted=np.std(filter_outliers(kp), ddof=1),
+            keypresses_standard_deviation_adjusted=np.std(my_filter_outliers(kp), ddof=1),
 
             sliderend_release_times=se,
             sliderend_release_median=np.median(se),
-            sliderend_release_median_adjusted=np.median(filter_outliers(se)),
+            sliderend_release_median_adjusted=np.median(my_filter_outliers(se)),
             sliderend_release_standard_deviation=np.std(se, ddof=1),
-            sliderend_release_standard_deviation_adjusted=np.std(filter_outliers(se), ddof=1),
+            sliderend_release_standard_deviation_adjusted=np.std(my_filter_outliers(se), ddof=1),
 
             judgements=judgements
         )