Merge pull request pennlabs#597 from pennlabs/pcx-history-probabilities

PCX history probabilities for PDP usage

backend/courses/management/commands/recompute_topics.py

@@ -1,9 +1,10 @@
 from django.core.management.base import BaseCommand
 from django.db import transaction
 from django.db.models import Count, OuterRef, Subquery
+from tqdm import tqdm
 
 from courses.models import Course, Topic
-from courses.util import all_semesters
+from courses.util import all_semesters, historical_semester_probability
 
 
 def garbage_collect_topics():
@@ -151,5 +152,28 @@ def handle(self, *args, **kwargs):
             assert (
                 min_semester in all_semesters()
             ), f"--min-semester={min_semester} is not a valid semester."
-
+        semesters = sorted(
+            [sem for sem in all_semesters() if not min_semester or sem >= min_semester]
+        )
         recompute_topics(min_semester, verbose=True, allow_null_parent_topic=bool(min_semester))
+        recompute_historical_semester_probabilities(current_semester=semesters[-1], verbose=True)
+
+
+def recompute_historical_semester_probabilities(current_semester, verbose=False):
+    """
+    Recomputes the historical probabilities for all topics.
+    """
+    if verbose:
+        print("Recomputing historical probabilities for all topics...")
+    topics = Topic.objects.all()
+    # Iterate over each Topic
+    for i, topic in tqdm(enumerate(topics), disable=not verbose, total=topics.count()):
+        # Calculate historical_year_probability for the current topic
+        ordered_courses = topic.courses.all().order_by("semester")
+        ordered_semester = [course.semester for course in ordered_courses]
+        historical_prob = historical_semester_probability(current_semester, ordered_semester)
+        # Update the historical_probabilities field for the current topic
+        topic.historical_probabilities_spring = historical_prob[0]
+        topic.historical_probabilities_summer = historical_prob[1]
+        topic.historical_probabilities_fall = historical_prob[2]
+        topic.save()

backend/courses/migrations/0064_auto_20240225_1331.py

@@ -0,0 +1,66 @@
+# Generated by Django 3.2.23 on 2024-02-25 18:31
+
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ("courses", "0063_auto_20231212_1750"),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name="topic",
+            name="historical_probabilities_fall",
+            field=models.FloatField(
+                default=0,
+                help_text="\nThe historical probability of a student taking a course in this topic in the fall\nsemester, based on historical data. This field is recomputed nightly from the\n`parent_course` graph (in the recompute_soft_state cron job).\n",
+            ),
+        ),
+        migrations.AddField(
+            model_name="topic",
+            name="historical_probabilities_spring",
+            field=models.FloatField(
+                default=0,
+                help_text="\nThe historical probability of a student taking a course in this topic in the spring\nsemester, based on historical data. This field is recomputed nightly from the\n`parent_course` graph (in the recompute_soft_state cron job).\n",
+            ),
+        ),
+        migrations.AddField(
+            model_name="topic",
+            name="historical_probabilities_summer",
+            field=models.FloatField(
+                default=0,
+                help_text="\nThe historical probability of a student taking a course in this topic in the summer\nsemester, based on historical data. This field is recomputed nightly from the\n`parent_course` graph (in the recompute_soft_state cron job).\n",
+            ),
+        ),
+        migrations.AlterField(
+            model_name="section",
+            name="activity",
+            field=models.CharField(
+                choices=[
+                    ("", "Undefined"),
+                    ("CLN", "Clinic"),
+                    ("CRT", "Clinical Rotation"),
+                    ("DAB", "Dissertation Abroad"),
+                    ("DIS", "Dissertation"),
+                    ("DPC", "Doctoral Program Exchange"),
+                    ("FLD", "Field Work"),
+                    ("HYB", "Hybrid"),
+                    ("IND", "Independent Study"),
+                    ("LAB", "Lab"),
+                    ("LEC", "Lecture"),
+                    ("MST", "Masters Thesis"),
+                    ("ONL", "Online"),
+                    ("PRC", "Practicum"),
+                    ("REC", "Recitation"),
+                    ("SEM", "Seminar"),
+                    ("SRT", "Senior Thesis"),
+                    ("STU", "Studio"),
+                ],
+                db_index=True,
+                help_text='The section activity, e.g. `LEC` for CIS-120-001 (2020A). Options and meanings: <table width=100%><tr><td>""</td><td>"Undefined"</td></tr><tr><td>"CLN"</td><td>"Clinic"</td></tr><tr><td>"CRT"</td><td>"Clinical Rotation"</td></tr><tr><td>"DAB"</td><td>"Dissertation Abroad"</td></tr><tr><td>"DIS"</td><td>"Dissertation"</td></tr><tr><td>"DPC"</td><td>"Doctoral Program Exchange"</td></tr><tr><td>"FLD"</td><td>"Field Work"</td></tr><tr><td>"HYB"</td><td>"Hybrid"</td></tr><tr><td>"IND"</td><td>"Independent Study"</td></tr><tr><td>"LAB"</td><td>"Lab"</td></tr><tr><td>"LEC"</td><td>"Lecture"</td></tr><tr><td>"MST"</td><td>"Masters Thesis"</td></tr><tr><td>"ONL"</td><td>"Online"</td></tr><tr><td>"PRC"</td><td>"Practicum"</td></tr><tr><td>"REC"</td><td>"Recitation"</td></tr><tr><td>"SEM"</td><td>"Seminar"</td></tr><tr><td>"SRT"</td><td>"Senior Thesis"</td></tr><tr><td>"STU"</td><td>"Studio"</td></tr></table>',
+                max_length=50,
+            ),
+        ),
+    ]

backend/courses/models.py

@@ -377,6 +377,36 @@ class Topic(models.Model):
         ),
     )
 
+    historical_probabilities_spring = models.FloatField(
+        default=0,
+        help_text=dedent(
+            """
+        The historical probability of a student taking a course in this topic in the spring
+        semester, based on historical data. This field is recomputed nightly from the
+        `parent_course` graph (in the recompute_soft_state cron job).
+        """
+        ),
+    )
+    historical_probabilities_summer = models.FloatField(
+        default=0,
+        help_text=dedent(
+            """
+        The historical probability of a student taking a course in this topic in the summer
+        semester, based on historical data. This field is recomputed nightly from the
+        `parent_course` graph (in the recompute_soft_state cron job).
+        """
+        ),
+    )
+    historical_probabilities_fall = models.FloatField(
+        default=0,
+        help_text=dedent(
+            """
+        The historical probability of a student taking a course in this topic in the fall
+        semester, based on historical data. This field is recomputed nightly from the
+        `parent_course` graph (in the recompute_soft_state cron job).
+        """
+        ),
+    )
     branched_from = models.ForeignKey(
         "Topic",
         related_name="branched_to",

backend/courses/util.py

@@ -721,3 +721,53 @@ def get_semesters(semesters: str = None) -> list[str]:
             if s not in possible_semesters:
                 raise ValueError(f"Provided semester {s} was not found in the db.")
     return sorted(semesters)
+
+
+def historical_semester_probability(current_semester: str, semesters: list[str]):
+    """
+    :param current: The current semester represented in the 20XX(A|B|C) format.
+    :type current: str
+    :param courses: A list of Course objects sorted by date in ascending order.
+    :type courses: list
+    :returns: A list of 3 probabilities representing the likelihood of
+    taking a course in each semester.
+    :rtype: list
+    """
+    PROB_DISTRIBUTION = [0.4, 0.3, 0.15, 0.1, 0.05]
+
+    def normalize_and_round(prob, i):
+        """Modifies the probability distribution to account for the
+        fact that the last course was taken i years ago."""
+        truncate = PROB_DISTRIBUTION[:i]
+        total = sum(truncate)
+        return list(map(lambda x: round(x / total, 3), truncate))
+
+    semester_probabilities = {"A": 0.0, "B": 0.0, "C": 0.0}
+    current_year = int(current_semester[:-1])
+    semesters = [
+        semester
+        for semester in semesters
+        if semester < str(current_year) and semester > str(current_year - 5)
+    ]
+    if not semesters:
+        return [0, 0, 0]
+    if current_year - int(semesters[0][:-1]) < 5:
+        # If the class hasn't been offered in the last 5 years,
+        # we make sure the resulting probabilities sum to 1
+        modified_prob_distribution = normalize_and_round(
+            PROB_DISTRIBUTION, current_year - int(semesters[0][:-1])
+        )
+    else:
+        modified_prob_distribution = PROB_DISTRIBUTION
+    for historical_semester in semesters:
+        historical_year = int(historical_semester[:-1])
+        sem_char = historical_semester[-1].upper()  # A, B, C
+        semester_probabilities[sem_char] += modified_prob_distribution[
+            current_year - historical_year - 1
+        ]
+    return list(
+        map(
+            lambda x: min(round(x, 2), 1.00),
+            [semester_probabilities["A"], semester_probabilities["B"], semester_probabilities["C"]],
+        )
+    )