Index: backend/subjects/consts.py =================================================================== --- backend/subjects/consts.py (revision b9a043ae3cd897c768eb2bbfc70479463ea55ec4) +++ backend/subjects/consts.py (revision 5dd5fcb6767602ba71469766386a300613beee10) @@ -3,8 +3,8 @@ WEIGHTS = { - "professors": 0.0325, - "assistants": 0.0325, - "technologies": 0.065, - "tags": 0.4, + "professors": 0.055, + "assistants": 0.055, + "technologies": 0.075, + "tags": 0.35, "evaluation": 0.15, "effort": 0.3, Index: backend/subjects/utils.py =================================================================== --- backend/subjects/utils.py (revision b9a043ae3cd897c768eb2bbfc70479463ea55ec4) +++ backend/subjects/utils.py (revision 5dd5fcb6767602ba71469766386a300613beee10) @@ -258,22 +258,94 @@ return subjects_tag_scores -def get_recommended_subjects(subjects_tag_scores): - """ - generates a list of recommended subjects based on weighted scores. - - args: - filtered_subjects_vector (dict): a dictionary where each key is a subject and each value is another dictionary - mapping feature keys to their corresponding scores for that subject. - returns: - list: a list of top N subject names recommended based on their normalized scores. if all scores are zero, returns empty list. - """ - subject_scores = {} - for subject in subjects_tag_scores: - keys = subjects_tag_scores[subject] - score = 0 - for key in keys: - score += WEIGHTS[key] * keys[key] - subject_scores[subject] = score - - top_subjects = list(dict(sorted(subject_scores.items(), key=lambda item: item[1], reverse=True)))[:NUMBER_OF_SUGGESTIONS] - return top_subjects +def get_explanation_message(criterion, score): + """Generates a human-readable explanation for a single matching criterion.""" + + # Thresholds to decide if a match is significant enough to be an "explanation" + thresholds = { + 'tags': 0.7, 'evaluation': 0.5, 'technologies': 0.5, + 'professors': 0.5, 'assistants': 0.5, 'participant_score': 0.5, + } + + if score < thresholds.get(criterion, 1.0): + return None + + messages = { + 'tags': f"Супер совпаѓање со твоите полиња на интерес ({score:.1%})", + 'evaluation': f"Се совпаѓа со твоите посакувани методи на евалуација ({score:.1%})", + 'technologies': f"Се совпаѓа со технологиите кои ги сакаш ({score:.1%})", + 'professors': f"Го предаваат професори кои ги сакаш ({score:.1%})", + 'assistants': f"Има асистенти кои ги сакаш ({score:.1%})", + 'participant_score': f"Има голем број на студенти", + } + return messages.get(criterion) + +# def get_detailed_tag_matches(student_vector, subject_vector): +# """Identifies the specific tags that matched between the student and subject.""" + +# student_tags_indices = {i for i, val in enumerate(student_vector['tags']) if val == 1} +# subject_tags_indices = {i for i, val in enumerate(subject_vector['tags']) if val == 1} + +# matching_indices = student_tags_indices.intersection(subject_tags_indices) + +# # Map indices back to tag names from the vocabulary +# all_tags = VOCABULARY.get('tags', []) +# matching_tags = [all_tags[i] for i in matching_indices if i < len(all_tags)] + +# return matching_tags + + +def get_recommendations_with_details(subjects_tag_scores): + """ + Generates a sorted list of recommended subjects with detailed explanations. + + Args: + subjects_tag_scores (dict): Scores for each subject across different criteria. + + Returns: + list: A list of dictionaries, each containing detailed info for a recommended subject. + """ + detailed_results = [] + + for subject_name, individual_scores in subjects_tag_scores.items(): + total_score = 0 + weighted_scores = {} + explanations = [] + + for criterion, score in individual_scores.items(): + weight = WEIGHTS.get(criterion, 0) + weighted_score = weight * score + total_score += weighted_score + weighted_scores[criterion] = weighted_score + + message = get_explanation_message(criterion, score) + if message: + explanations.append(message) + + # Sort explanations by their match percentage, it extracts the percentage from the message (e.g. Something (72.8%)) and sorts them in descending order + explanations.sort(key=lambda msg: -(float(msg.split('(')[-1].replace('%)', '')) if '(' in msg and '%' in msg else float('-inf'))) + # primary_reason_criterion = max(weighted_scores, key=weighted_scores.get) + # primary_reason_score = individual_scores[primary_reason_criterion] + # primary_reason = get_explanation_message(primary_reason_criterion, primary_reason_score) + # if not primary_reason: + # primary_reason = f"Добро совпаѓање поради {primary_reason_criterion.replace('_', ' ')}" + + # matching_tags = get_detailed_tag_matches(student_vector, eligible_subjects_dict[subject_name]) + + detailed_results.append({ + 'subject_name': subject_name, + 'total_score': total_score, + # 'primary_reason': primary_reason, + 'explanations': explanations, + # 'detailed_scores': individual_scores, + # 'matching_tags': matching_tags, + 'match_percentage': min(total_score * 100, 100) + }) + + + max_score = max([res['total_score'] for res in detailed_results]) or 1 + for res in detailed_results: + res['match_percentage'] = round((res['total_score'] / max_score) * 100, 1) + + detailed_results.sort(key=lambda x: x['total_score'], reverse=True) + return detailed_results[:NUMBER_OF_SUGGESTIONS] + Index: backend/subjects/views.py =================================================================== --- backend/subjects/views.py (revision b9a043ae3cd897c768eb2bbfc70479463ea55ec4) +++ backend/subjects/views.py (revision 5dd5fcb6767602ba71469766386a300613beee10) @@ -9,9 +9,10 @@ from rest_framework.permissions import IsAuthenticated from django.db.models import Case, When, Count, F, Q -from subjects.utils import get_eligible_subjects, get_recommendations_cache_key, get_recommended_subjects, map_to_subjects_vector, score_for_preferences, get_student_vector +from subjects.utils import get_eligible_subjects, get_recommendations_cache_key, get_recommendations_with_details, map_to_subjects_vector, score_for_preferences, get_student_vector from .serializers import SubjectSerializer, EvaluationReviewSerializer, OtherReviewSerializer from .models import Subject, Review, EvaluationReview, OtherReview, ReviewVote from rest_framework.pagination import LimitOffsetPagination from auth_form.permissions import IsStudent, IsAdmin +import logging def index(request): @@ -42,22 +43,49 @@ cached_data = cache.get(cache_key) if cached_data: - return Response({"data": json.loads(cached_data)}, status=status.HTTP_200_OK) - try: - subjects = get_eligible_subjects(student, season=season, not_activated=not_activated) - subject_vectors = map_to_subjects_vector(subjects) + return Response(json.loads(cached_data), status=status.HTTP_200_OK) + try: + eligible_subjects = get_eligible_subjects(student, season=season, not_activated=not_activated) + if not eligible_subjects: + return Response({"data": []}, status=status.HTTP_200_OK) + + eligible_subjects_dict = map_to_subjects_vector(eligible_subjects) student_vector = get_student_vector(student) - final_subjects = get_recommended_subjects(score_for_preferences(student_vector, subject_vectors)) - - order = Case(*[When(name=subject_name, then=pos) for pos, subject_name in enumerate(final_subjects)]) - - recommended_subject_objects = Subject.objects.filter(name__in=final_subjects).order_by(order) + subjects_scores = score_for_preferences(student_vector, eligible_subjects_dict) + + recommendations_with_details = get_recommendations_with_details( + subjects_scores + ) + + if not recommendations_with_details: + return Response({"data": []}, status=status.HTTP_200_OK) + final_subject_names = [rec['subject_name'] for rec in recommendations_with_details] + details_map = {rec['subject_name']: rec for rec in recommendations_with_details} + + order = Case(*[When(name=name, then=pos) for pos, name in enumerate(final_subject_names)]) + recommended_subject_objects = Subject.objects.filter(name__in=final_subject_names).order_by(order) serializer = SubjectSerializer(recommended_subject_objects, many=True) + + final_response_data = [] + for subject_data in serializer.data: + details = details_map.get(subject_data['name']) + if details: + subject_data['recommendation_details'] = { + 'match_percentage': details['match_percentage'], + # 'primary_reason': details['primary_reason'], + 'explanations': details['explanations'], + # 'matching_tags': details['matching_tags'], + # 'detailed_scores': details['detailed_scores'] + } + final_response_data.append(subject_data) + + response_payload = {"data": final_response_data} if cache_key: - cache.set(cache_key, json.dumps(serializer.data), timeout=60 * 60 * 24 * 14) # 14 days - return Response({"data": serializer.data}, status=status.HTTP_200_OK) + cache.set(cache_key, json.dumps(response_payload), timeout=60 * 60 * 24 * 14) + return Response(response_payload, status=status.HTTP_200_OK) except Exception as e: + logging.error(f"Recommendation error for student {student.id}: {e}", exc_info=True) return Response({"message": f"An error occurred: {str(e)}"}, status=status.HTTP_500_INTERNAL_SERVER_ERROR) Index: frontend/src/components/SubjectCatalog/SubjectCard.tsx =================================================================== --- frontend/src/components/SubjectCatalog/SubjectCard.tsx (revision b9a043ae3cd897c768eb2bbfc70479463ea55ec4) +++ frontend/src/components/SubjectCatalog/SubjectCard.tsx (revision 5dd5fcb6767602ba71469766386a300613beee10) @@ -12,4 +12,11 @@ isRecommended?: boolean; isLoading?: boolean; + recommendationDetails?: { + match_percentage: number; + primary_reason: string; + explanations: string[]; + matching_tags: string[]; + detailed_scores: Record; + }; } @@ -22,35 +29,82 @@ isRecommended = false, isLoading = false, + recommendationDetails, }: SubjectCardProps) => { return (
-
-
-
-

{subject.name}

-

{subject.code}

+
+
+
+

+ {subject.name} +

+

{subject.code}

-
- {isRecommended && subject.subject_info.activated === false ? ( - + + {isRecommended && subject.subject_info.activated === false && ( +
+ Никогаш не бил активиран! - ) : isFirst ? ( - - Најсоодветен! - - ) : null} +
+ )} + +
+ {recommendationDetails && ( +
50 + ? "bg-blue-50 text-blue-800 border-blue-200" + : "bg-red-50 text-red-800 border-red-200" + }`} + > + {isFirst ? ( +
+ + Најсоодветен! + +
+ ) : ( +

+ {recommendationDetails.match_percentage >= 90 + ? "Супер за тебе" + : recommendationDetails.match_percentage >= 75 + ? "Многу добар избор" + : recommendationDetails.match_percentage >= 50 + ? "Добар избор" + : "Не е најдобриот избор"} +

+ )} +
    + {recommendationDetails.explanations.map( + (explanation, index) => ( +
  • + • {explanation} +
    • + ) + )} +
+
+ )}
-
+ +
- @@ -62,5 +116,5 @@ )}
-
+
Index: frontend/src/components/types.ts =================================================================== --- frontend/src/components/types.ts (revision b9a043ae3cd897c768eb2bbfc70479463ea55ec4) +++ frontend/src/components/types.ts (revision 5dd5fcb6767602ba71469766386a300613beee10) @@ -15,4 +15,11 @@ abstract: string; subject_info: SubjectInfo; + recommendation_details?: { + match_percentage: number; + primary_reason: string; + explanations: string[]; + matching_tags: string[]; + detailed_scores: Record; + }; } Index: frontend/src/pages/Recommendations.tsx =================================================================== --- frontend/src/pages/Recommendations.tsx (revision b9a043ae3cd897c768eb2bbfc70479463ea55ec4) +++ frontend/src/pages/Recommendations.tsx (revision 5dd5fcb6767602ba71469766386a300613beee10) @@ -184,5 +184,5 @@

Вашите препораки за {getSeasonText().toLowerCase()}{" "} - {getSeasonText().toLowerCase() === "Зимски + Летен" + {getSeasonText().toLowerCase() === "зимски + летен" ? "семестри" : "семестар"} @@ -193,12 +193,10 @@

-
+
{recommendations.map((subject, index) => (