noa151/LeetCodePredictions
0
1import json
2import gradio as gr
3import joblib
4import pandas as pd
5from related_topics_prediction import MultiLabelThresholdOptimizer
6
7
8def convert_to_float(value):
9 if 'K' in value:
10 return float(value.replace('K', '')) * 1_000
11 elif 'M' in value:
12 return float(value.replace('M', '')) * 1_000_000
13 return float(value) # If it's already a number
14
15
16def convert_to_string(value):
17 if value >= 1_000_000:
18 return f"{value / 1_000_000:.1f}M"
19 elif value >= 1_000:
20 return f"{value / 1_000:.1f}K"
21 return str(int(value)) # Keep it as an integer if it's below 1,000
22
23
24def greet(title, description, difficulty, topics, likes, accepted, submission, comments, is_premium, predict):
25
26 x_new = pd.DataFrame([{
27 'id': 1,
28 'title': str(title),
29 'description': str(description),
30 'is_premium': 1 if is_premium == "premium" else 0,
31 'difficulty': 0 if difficulty == "Easy" else 1 if difficulty == "Hard" else 2,
32 'acceptance_rate': convert_to_float(accepted)/convert_to_float(submission),
33 'frequency': 0,
34 'discuss_count': float(comments),
35 'accepted': convert_to_float(accepted),
36 'submissions': convert_to_float(submission),
37 'companies': [""],
38 'related_topics': topics.split(',') if isinstance(topics, str) else topics,
39 'likes': convert_to_float(likes),
40 'dislikes': 0,
41 'rating': convert_to_float(likes) / (convert_to_float(likes) + 0),
42 'asked_by_faang': 0,
43 'similar_questions': ""
44 }])
45
46 # Efficient Multi-Hot Encoding for Companies
47 company_data = {company: 1 if company in x_new["companies"].iloc[0] else 0 for company in companies_columns}
48 x_new = pd.concat([x_new, pd.DataFrame([company_data])], axis=1)
49
50 x_new = x_new.drop(columns=["companies"]) # Drop original column
51
52 # Efficient Multi-Hot Encoding for Topics
53 topic_data = {topic: 1 if topic in x_new["related_topics"].iloc[0] else 0 for topic in the_topics}
54 x_new = pd.concat([x_new, pd.DataFrame([topic_data])], axis=1)
55
56 x_new = x_new.drop(columns=["related_topics"]) # Drop original topics column
57
58 # Label encode 'title'
59 title_model = joblib.load("title_encoder.pkl")
60 x_new['title'] = title_model.fit_transform(x_new['title'])
61
62 if predict == "related topics":
63 vectorizer = joblib.load("related_topics_vectorizer.pkl")
64
65 new_tfidf = vectorizer.transform(x_new["description"])
66
67 best_model_info = joblib.load('best_model_related_topics_info.pkl')
68 best_model = joblib.load("best_related_topics_model.pkl")
69 optimizer = MultiLabelThresholdOptimizer()
70 optimizer.optimal_thresholds[best_model_info['model_name']] = best_model_info['threshold']
71
72 predictions = optimizer.predict(best_model, new_tfidf, best_model_info['model_name'])
73
74 mlb = joblib.load("related_topics_label_binarizer.pkl")
75 predictions = mlb.inverse_transform(predictions)
76
77 ans = f"the related topics are: {', '.join(map(str, predictions[0]))}"
78 return ans
79
80 else:
81 vectorizer = joblib.load("tfidf_vectorizer.pkl")
82
83 new_tfidf = vectorizer.transform(x_new["description"])
84
85 # Convert to DataFrame
86 new_tfidf_df = pd.DataFrame(new_tfidf.toarray(), columns=vectorizer.get_feature_names_out())
87 x_new = pd.concat([x_new, new_tfidf_df], axis=1)
88 x_new = x_new.drop(columns=['description'])
89
90 if predict == "difficulty level":
91 # load the dislike model because there is no dislike in the input
92 dislikes_model, feature_names = joblib.load("dislikes_XGB_regression_model.pkl")
93
94 x_new_filtered = x_new[feature_names] # Select only the required features
95 dislike = dislikes_model.predict(x_new_filtered)
96 x_new['dislikes'] = dislike[0]
97 x_new['rating']: convert_to_float(likes) / (convert_to_float(likes) + dislike[0])
98
99 # Load the model
100 class_model = joblib.load("level_classifier_model.pkl")
101
102 # Get feature names from trained model
103 trained_feature_names = class_model.named_steps['standardscaler'].get_feature_names_out()
104
105 x_new = x_new[trained_feature_names] # Reorder and remove extra columns
106
107 # Fill missing columns with 0 (or a suitable default)
108 for col in trained_feature_names:
109 if col not in x_new:
110 x_new[col] = 0 # or another default value
111
112 x_new = x_new[trained_feature_names] # Ensure correct order again
113
114 predictions = class_model.predict(x_new)
115
116 if predictions == 1:
117 prediction = "Hard"
118 elif predictions == 0:
119 prediction = "Easy"
120 elif predictions == 2:
121 prediction = "Medium"
122
123 ans = f"the level difficulty is: {prediction}"
124 return ans
125
126 elif predict == "acceptance":
127 # Load the model
128 accepted_submissions_model, feature_names = joblib.load("accepted_submissions_regression_model.pkl")
129
130 # Assuming `X_new` is a DataFrame with extra features
131 x_new_filtered = x_new[feature_names] # Select only the required features
132
133 predictions = accepted_submissions_model.predict(x_new_filtered)
134
135 ans = f"the accepted is: {convert_to_string(predictions[0])}"
136 return ans
137
138 elif predict == "number of likes":
139 # Load the model
140 likes_model, feature_names = joblib.load("likes_random_forest_regression_model.pkl")
141
142 # Assuming `X_new` is a DataFrame with extra features
143 x_new_filtered = x_new[feature_names] # Select only the required features
144
145 predictions = likes_model.predict(x_new_filtered)
146
147 ans = f"the likes amount is: {convert_to_string(predictions[0])}"
148 return ans
149
150 elif predict == "number of dislikes":
151 # Load the model
152 dislikes_model, feature_names = joblib.load("dislikes_XGB_regression_model.pkl")
153
154 # Assuming `x_new` is a DataFrame with extra features
155 x_new_filtered = x_new[feature_names] # Select only the required features
156
157 predictions = dislikes_model.predict(x_new_filtered)
158
159 ans = f"the dislikes amount is: {convert_to_string(predictions[0])}"
160 return ans
161
162
163with open("encoding_metadata.json", "r") as f:
164 encoding_metadata = json.load(f)
165
166the_topics = encoding_metadata["related_topics_columns"]
167the_topics.remove("")
168companies_columns = encoding_metadata["companies_columns"]
169companies_columns.remove("")
170
171demo = gr.Interface(
172 fn=greet,
173 inputs=[gr.Text(label="Title"), gr.Text(label="Description"),
174 gr.Radio(choices=["Easy", "Medium", "Hard"], label="Difficulty Level"),
175 gr.Dropdown(the_topics, multiselect=True, label="Related Topics",
176 info="choose all the related topics of this question"),
177 gr.Text(label="Likes Amount"),
178 gr.Text(label="Accepted Amount"),
179 gr.Text(label="Submission Amount"),
180 gr.Text(label="Comments Amount"),
181 gr.Radio(choices=["premium", "not premium"], label="Is Premium"),
182 gr.Radio(choices=["acceptance", "difficulty level", "number of likes", "number of dislikes",
183 "related topics"], label="Please Predict..")
184 ],
185 outputs=[gr.Text(label="The Prediction")],
186 title="LEETCODE PREDICTOR",
187 description="please go to the leetcode website (https://leetcode.com/problemset/) choose a question and copy the question's detiles to the relevant spaces, then choose what you whould like to predict and submit. the prediction result will appear on the right side of the screen 😉"
188)
189
190demo.launch()
191