![\"\"](\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2024\/11\/ca5cfa60-0ff6-46a2-be52-629f7ae0b05a-300x167-1.png\")
<\/p>\n<\/p>\n
\u4ee3\u7801\u751f\u6210\u548c\u8bc4\u4f30\u6700\u8fd1\u5df2\u7ecf\u6210\u4e3a\u8bb8\u591a\u5546\u4e1a\u4ea7\u54c1\u63d0\u4f9b\u7684\u91cd\u8981\u529f\u80fd\uff0c\u4ee5\u5e2e\u52a9\u5f00\u53d1\u4eba\u5458\u5904\u7406\u4ee3\u7801\u3002LLM\u8fd8\u53ef\u4ee5\u8fdb\u4e00\u6b65\u7528\u4e8e\u5904\u7406\u6570\u636e\u79d1\u5b66\u5de5\u4f5c\uff0c\u5c24\u5176\u662f\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u3002\u672c\u6587\u5c06\u63a2\u8ba8\u5982\u4f55\u5c06\u81ea\u52a8\u5316\u7528\u4e8e\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u3002<\/p>\n
<\/p>\n
\u5927\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u5df2\u6210\u4e3a\u4e00\u79cd\u5de5\u5177\uff0c\u4ece\u56de\u7b54\u95ee\u9898\u5230\u751f\u6210\u4efb\u52a1\u5217\u8868\uff0c\u5b83\u4eec\u5728\u8bb8\u591a\u65b9\u9762\u7b80\u5316\u4e86\u6211\u4eec\u7684\u5de5\u4f5c\u3002\u5982\u4eca\u4e2a\u4eba\u548c\u4f01\u4e1a\u5df2\u7ecf\u4f7f\u7528LLM\u6765\u5e2e\u52a9\u5b8c\u6210\u5de5\u4f5c\u3002<\/p>\n
\u4ee3\u7801\u751f\u6210\u548c\u8bc4\u4f30\u6700\u8fd1\u5df2\u7ecf\u6210\u4e3a\u8bb8\u591a\u5546\u4e1a\u4ea7\u54c1\u63d0\u4f9b\u7684\u91cd\u8981\u529f\u80fd\uff0c\u4ee5\u5e2e\u52a9\u5f00\u53d1\u4eba\u5458\u5904\u7406\u4ee3\u7801\u3002LLM\u8fd8\u53ef\u4ee5\u8fdb\u4e00\u6b65\u7528\u4e8e\u5904\u7406\u6570\u636e\u79d1\u5b66\u5de5\u4f5c\uff0c\u5c24\u5176\u662f\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u3002<\/p>\n
\u672c\u6587\u5c06\u63a2\u8ba8\u5982\u4f55\u5c06\u81ea\u52a8\u5316\u7528\u4e8e\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u3002<\/p>\n
\u6211\u4eec\u5c06\u8bbe\u7f6e\u7528\u4e8e\u6a21\u578b\u8bad\u7ec3\u7684\u6570\u636e\u96c6\u548c\u7528\u4e8e\u81ea\u52a8\u5316\u7684\u4ee3\u7801\u3002\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528\u6765\u81eaKaggle\u7684\u4fe1\u7528\u6c7d\u8f66\u6b3a\u8bc8\u6570\u636e\u96c6\u3002\u4ee5\u4e0b\u662f\u6211\u4e3a\u9884\u5904\u7406\u8fc7\u7a0b\u6240\u505a\u7684\u51c6\u5907\u3002<\/p>\n
\u590d\u5236<\/p>\n
import pandas as pd df = pd.read_csv(‘fraud_data.csv’) df = df.drop([‘trans_date_trans_time’, ‘merchant’, ‘dob’, ‘trans_num’, ‘merch_lat’, ‘merch_long’], axis =1) df = df.dropna().reset_index(drop = True) df.to_csv(‘fraud_data.csv’, index = False)<\/p>\n
\u6211\u4eec\u5c06\u53ea\u4f7f\u7528\u4e00\u4e9b\u6570\u636e\u96c6\uff0c\u4e22\u5f03\u6240\u6709\u7f3a\u5931\u7684\u6570\u636e\u3002\u8fd9\u4e0d\u662f\u6700\u4f18\u7684\u8fc7\u7a0b\uff0c\u4f46\u6211\u4eec\u5173\u6ce8\u7684\u662f\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u3002<\/p>\n
\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u5c06\u4e3a\u6211\u4eec\u7684\u9879\u76ee\u51c6\u5907\u4e00\u4e2a\u6587\u4ef6\u5939\uff0c\u5c06\u6240\u6709\u76f8\u5173\u6587\u4ef6\u653e\u5728\u90a3\u91cc\u3002\u9996\u5148\uff0c\u6211\u4eec\u5c06\u4e3a\u73af\u5883\u521b\u5efarequirements.txt\u6587\u4ef6\u3002\u4f60\u53ef\u4ee5\u7528\u4e0b\u9762\u7684\u8f6f\u4ef6\u5305\u6765\u586b\u5145\u5b83\u4eec\u3002<\/p>\n
\u590d\u5236<\/p>\n
openai pandas scikit-learn pyyaml<\/p>\n
\u63a5\u4e0b\u6765\uff0c\u6211\u4eec\u5c06\u4e3a\u6240\u6709\u76f8\u5173\u7684\u5143\u6570\u636e\u4f7f\u7528YAML\u6587\u4ef6\u3002\u8fd9\u5c06\u5305\u62ecOpenAI API\u5bc6\u94a5\u3001\u8981\u6d4b\u8bd5\u7684\u6a21\u578b\u3001\u8bc4\u4f30\u5ea6\u91cf\u6307\u6807\u548c\u6570\u636e\u96c6\u7684\u4f4d\u7f6e\u3002<\/p>\n
\u590d\u5236<\/p>\n
llm_api_key: “YOUR-OPENAI-API-KEY” default_models: – LogisticRegression – DecisionTreeClassifier – RandomForestClassifier metrics: [“accuracy”, “precision”, “recall”, “f1_score”] dataset_path: “fraud_data.csv”<\/p>\n
\u7136\u540e\uff0c\u6211\u4eec\u5bfc\u5165\u8fd9\u4e2a\u8fc7\u7a0b\u4e2d\u4f7f\u7528\u7684\u8f6f\u4ef6\u5305\u3002\u6211\u4eec\u5c06\u4f9d\u9760Scikit-Learn\u7528\u4e8e\u5efa\u6a21\u8fc7\u7a0b\uff0c\u5e76\u4f7f\u7528OpenAI\u7684GPT-4\u4f5c\u4e3aLLM\u3002<\/p>\n
\u590d\u5236<\/p>\n
import pandas as pd import yaml import ast import re import sklearn from openai import OpenAI from sklearn.linear_model import LogisticRegression from sklearn.tree import DecisionTreeClassifier from sklearn.ensemble import RandomForestClassifier from sklearn.model_selection import train_test_split from sklearn.preprocessing import LabelEncoder from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score<\/p>\n
\u6b64\u5916\uff0c\u6211\u4eec\u5c06\u8bbe\u7f6e\u8f85\u52a9\uff08helper\uff09\u51fd\u6570\u548c\u4fe1\u606f\u6765\u5e2e\u52a9\u8be5\u8fc7\u7a0b\u3002\u4ece\u6570\u636e\u96c6\u52a0\u8f7d\u5230\u6570\u636e\u9884\u5904\u7406\uff0c\u914d\u7f6e\u52a0\u8f7d\u5668\u5728\u5982\u4e0b\u7684\u51fd\u6570\u4e2d\u3002<\/p>\n
\u590d\u5236<\/p>\n
model_mapping = { “LogisticRegression”: LogisticRegression, “DecisionTreeClassifier”: DecisionTreeClassifier, “RandomForestClassifier”: RandomForestClassifier } def load_config(config_path=’config.yaml’): with open(config_path, ‘r’) as file: config = yaml.safe_load(file) return config def load_data(dataset_path): return pd.read_csv(dataset_path) def preprocess_data(df): label_encoders = {} for column in df.select_dtypes(include=[‘object’]).columns: le = LabelEncoder() df[column] = le.fit_transform(df[column]) label_encoders[column] = le return df, label_encoders<\/p>\n
\u5728\u540c\u4e00\u4e2a\u6587\u4ef6\u4e2d\uff0c\u6211\u4eec\u5c06LLM\u8bbe\u7f6e\u4e3a\u626e\u6f14\u673a\u5668\u5b66\u4e60\u89d2\u8272\u7684\u4e13\u5bb6\u3002\u6211\u4eec\u5c06\u4f7f\u7528\u4e0b\u9762\u7684\u4ee3\u7801\u6765\u542f\u52a8\u5b83\u3002<\/p>\n
\u590d\u5236<\/p>\n
def call_llm(prompt, api_key): client = OpenAI(api_key=api_key) response = client.chat.completions.create( model=”gpt-4″, messages=[ {“role”: “system”, “content”: “You are an expert in machine learning and able to evaluate the model well.”}, {“role”: “user”, “content”: prompt} ] ) return response.choices[0].message.content.strip()<\/p>\n
\u4f60\u53ef\u4ee5\u5c06LLM\u6a21\u578b\u66f4\u6539\u4e3a\u6240\u9700\u7684\u6a21\u578b\uff0c\u6bd4\u5982\u6765\u81eaHugging\u00a0Face\u7684\u5f00\u6e90\u6a21\u578b\uff0c\u4f46\u6211\u4eec\u5efa\u8bae\u6682\u4e14\u575a\u6301\u4f7f\u7528OpenAI\u3002<\/p>\n
\u6211\u5c06\u5728\u4e0b\u9762\u7684\u4ee3\u7801\u4e2d\u51c6\u5907\u4e00\u4e2a\u51fd\u6570\u6765\u6e05\u7406LLM\u7ed3\u679c\u3002\u8fd9\u786e\u4fdd\u4e86\u8f93\u51fa\u53ef\u4ee5\u7528\u4e8e\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u6b65\u9aa4\u7684\u540e\u7eed\u8fc7\u7a0b\u3002<\/p>\n
\u590d\u5236<\/p>\n
def clean_hyperparameter_suggestion(suggestion): pattern = r’\\{.*?\\}’ match = re.search(pattern, suggestion, re.DOTALL) if match: cleaned_suggestion = match.group(0) return cleaned_suggestion else: print(“Could not find a dictionary in the hyperparameter suggestion.”) return None def extract_model_name(llm_response, available_models): for model in available_models: pattern = r’\\b’ + re.escape(model) + r’\\b’ if re.search(pattern, llm_response, re.IGNORECASE): return model return None def validate_hyperparameters(model_class, hyperparameters): valid_params = model_class().get_params() invalid_params = [] for param, value in hyperparameters.items(): if param not in valid_params: invalid_params.append(param) else: if param == ‘max_features’ and value == ‘auto’: print(f”Invalid value for parameter ‘{param}’: ‘{value}'”) invalid_params.append(param) if invalid_params: print(f”Invalid hyperparameters for {model_class.__name__}: {invalid_params}”) return False return True def correct_hyperparameters(hyperparameters, model_name): corrected = False if model_name == “RandomForestClassifier”: if ‘max_features’ in hyperparameters and hyperparameters[‘max_features’] == ‘auto’: print(“Correcting ‘max_features’ from ‘auto’ to ‘sqrt’ for RandomForestClassifier.”) hyperparameters[‘max_features’] = ‘sqrt’ corrected = True return hyperparameters, corrected<\/p>\n
\u7136\u540e\uff0c\u6211\u4eec\u5c06\u9700\u8981\u8be5\u51fd\u6570\u6765\u542f\u52a8\u6a21\u578b\u548c\u8bc4\u4f30\u8bad\u7ec3\u8fc7\u7a0b\u3002\u4e0b\u9762\u7684\u4ee3\u7801\u5c06\u7528\u4e8e\u901a\u8fc7\u63a5\u53d7\u5206\u5272\u5668\u6570\u636e\u96c6\u3001\u6211\u4eec\u8981\u6620\u5c04\u7684\u6a21\u578b\u540d\u79f0\u4ee5\u53ca\u8d85\u53c2\u6570\u6765\u8bad\u7ec3\u6a21\u578b\u3002\u7ed3\u679c\u5c06\u662f\u5ea6\u91cf\u6307\u6807\u548c\u6a21\u578b\u5bf9\u8c61\u3002<\/p>\n
\u590d\u5236<\/p>\n
def train_and_evaluate(X_train, X_test, y_train, y_test, model_name, hyperparameters=None): if model_name not in model_mapping: print(f”Valid model names are: {list(model_mapping.keys())}”) return None, None model_class = model_mapping.get(model_name) try: if hyperparameters: hyperparameters, corrected = correct_hyperparameters(hyperparameters, model_name) if not validate_hyperparameters(model_class, hyperparameters): return None, None model = model_class(**hyperparameters) else: model = model_class() except Exception as e: print(f”Error instantiating model with hyperparameters: {e}”) return None, None try: model.fit(X_train, y_train) except Exception as e: print(f”Error during model fitting: {e}”) return None, None y_pred = model.predict(X_test) metrics = { “accuracy”: accuracy_score(y_test, y_pred), “precision”: precision_score(y_test, y_pred, average=’weighted’, zero_division=0), “recall”: recall_score(y_test, y_pred, average=’weighted’, zero_division=0), “f1_score”: f1_score(y_test, y_pred, average=’weighted’, zero_division=0) } return metrics, model<\/p>\n
\u51c6\u5907\u5c31\u7eea\u540e\uff0c\u6211\u4eec\u5c31\u53ef\u4ee5\u8bbe\u7f6e\u81ea\u52a8\u5316\u8fc7\u7a0b\u4e86\u3002\u6709\u51e0\u4e2a\u6b65\u9aa4\u6211\u4eec\u53ef\u4ee5\u5b9e\u73b0\u81ea\u52a8\u5316\uff0c\u5176\u4e2d\u5305\u62ec\uff1a<\/p>\n
1.\u8bad\u7ec3\u548c\u8bc4\u4f30\u6240\u6709\u6a21\u578b<\/p>\n
\u590d\u5236<\/p>\n
def run_llm_based_model_selection_experiment(df, config): #Model Training X = df.drop(“is_fraud”, axis=1) y = df[“is_fraud”] X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42) available_models = config[‘default_models’] model_performance = {} for model_name in available_models: print(f”Training model: {model_name}”) metrics, _ = train_and_evaluate(X_train, X_test, y_train, y_test, model_name) model_performance[model_name] = metrics print(f”Model: {model_name} | Metrics: {metrics}”) #LLM selecting the best model sklearn_version = sklearn.__version__ prompt = ( f”I have trained the following models with these metrics: {model_performance}. ” “Which model should I select based on the best performance?” ) best_model_response = call_llm(prompt, config[‘llm_api_key’]) print(f”LLM response for best model selection:\\n{best_model_response}”) best_model = extract_model_name(best_model_response, available_models) if not best_model: print(“Error: Could not extract a valid model name from LLM response.”) return print(f”LLM selected the best model: {best_model}”) #Check for hyperparameter tuning prompt_tuning = ( f”The selected model is {best_model}. Can you suggest hyperparameters for better performance? ” “Please provide them in Python dictionary format, like {‘max_depth’: 5, ‘min_samples_split’: 4}. ” f”Ensure that all suggested hyperparameters are valid for scikit-learn version {sklearn_version}, ” “and avoid using deprecated or invalid values such as ‘max_features’: ‘auto’. ” “Don’t provide any explanation or return in any other format.” ) tuning_suggestion = call_llm(prompt_tuning, config[‘llm_api_key’]) print(f”Hyperparameter tuning suggestion received:\\n{tuning_suggestion}”) cleaned_suggestion = clean_hyperparameter_suggestion(tuning_suggestion) if cleaned_suggestion is None: suggested_params = None else: try: suggested_params = ast.literal_eval(cleaned_suggestion) if not isinstance(suggested_params, dict): print(“Hyperparameter suggestion is not a valid dictionary.”) suggested_params = None except (ValueError, SyntaxError) as e: print(f”Error parsing hyperparameter suggestion: {e}”) suggested_params = None #Automatically run hyperparameter tuning if suggested if suggested_params: print(f”Running {best_model} with suggested hyperparameters: {suggested_params}”) tuned_metrics, _ = train_and_evaluate( X_train, X_test, y_train, y_test, best_model, hyperparameters=suggested_params ) print(f”Metrics after tuning: {tuned_metrics}”) else: print(“No valid hyperparameters were provided for tuning.”)<\/p>\n
\u5728\u4e0a\u9762\u7684\u4ee3\u7801\u4e2d\uff0c\u6211\u6307\u5b9a\u4e86LLM\u5982\u4f55\u6839\u636e\u8bd5\u9a8c\u8bc4\u4f30\u6211\u4eec\u7684\u6bcf\u4e2a\u6a21\u578b\u3002\u6211\u4eec\u4f7f\u7528\u4ee5\u4e0b\u63d0\u793a\u6839\u636e\u6a21\u578b\u7684\u6027\u80fd\u6765\u9009\u62e9\u8981\u4f7f\u7528\u7684\u6a21\u578b\u3002<\/p>\n
\u590d\u5236<\/p>\n
prompt = ( f”I have trained the following models with these metrics: {model_performance}. ” “Which model should I select based on the best performance?”)<\/p>\n
\u4f60\u59cb\u7ec8\u53ef\u4ee5\u66f4\u6539\u63d0\u793a\uff0c\u4ee5\u5b9e\u73b0\u6a21\u578b\u9009\u62e9\u7684\u4e0d\u540c\u89c4\u5219\u3002<\/p>\n
\u4e00\u65e6\u9009\u62e9\u4e86\u6700\u4f73\u6a21\u578b\uff0c\u6211\u5c06\u4f7f\u7528\u4ee5\u4e0b\u63d0\u793a\u6765\u5efa\u8bae\u5e94\u8be5\u4f7f\u7528\u54ea\u4e9b\u8d85\u53c2\u6570\u7528\u4e8e\u540e\u7eed\u8fc7\u7a0b\u3002\u6211\u8fd8\u6307\u5b9a\u4e86Scikit-Learn\u7248\u672c\uff0c\u56e0\u4e3a\u8d85\u53c2\u6570\u56e0\u7248\u672c\u7684\u4e0d\u540c\u800c\u6709\u53d8\u5316\u3002<\/p>\n
\u590d\u5236<\/p>\n
prompt_tuning = ( f”The selected model is {best_model}. Can you suggest hyperparameters for better performance? ” “Please provide them in Python dictionary format, like {‘max_depth’: 5, ‘min_samples_split’: 4}. ” f”Ensure that all suggested hyperparameters are valid for scikit-learn version {sklearn_version}, ” “and avoid using deprecated or invalid values such as ‘max_features’: ‘auto’. ” “Don’t provide any explanation or return in any other format.”)<\/p>\n
\u4f60\u53ef\u4ee5\u4ee5\u4efb\u4f55\u60f3\u8981\u7684\u65b9\u5f0f\u66f4\u6539\u63d0\u793a\uff0c\u6bd4\u5982\u901a\u8fc7\u66f4\u5927\u80c6\u5730\u5c1d\u8bd5\u8c03\u4f18\u8d85\u53c2\u6570\uff0c\u6216\u6dfb\u52a0\u53e6\u4e00\u79cd\u6280\u672f\u3002<\/p>\n
\u6211\u628a\u4e0a\u9762\u7684\u6240\u6709\u4ee3\u7801\u653e\u5728\u4e00\u4e2a\u540d\u4e3aautomated_model_llm.py\u7684\u6587\u4ef6\u4e2d\u3002\u6700\u540e\uff0c\u6dfb\u52a0\u4ee5\u4e0b\u4ee3\u7801\u4ee5\u8fd0\u884c\u6574\u4e2a\u8fc7\u7a0b\u3002<\/p>\n
\u590d\u5236<\/p>\n
def main(): config = load_config() df = load_data(config[‘dataset_path’]) df, _ = preprocess_data(df) run_llm_based_model_selection_experiment(df, config) if __name__ == “__main__”: main()<\/p>\n
\u4e00\u65e6\u4e00\u5207\u51c6\u5907\u5c31\u7eea\uff0c\u4f60\u5c31\u53ef\u4ee5\u8fd0\u884c\u4ee5\u4e0b\u4ee3\u7801\u6765\u6267\u884c\u4ee3\u7801\u3002<\/p>\n
\u590d\u5236<\/p>\n
python automated_model_llm.py<\/p>\n
\u8f93\u51fa\uff1a<\/p>\n
\u590d\u5236<\/p>\n
LLM selected the best model: RandomForestClassifier Hyperparameter tuning suggestion received: { ‘n_estimators’: 100, ‘max_depth’: None, ‘min_samples_split’: 2, ‘min_samples_leaf’: 1, ‘max_features’: ‘sqrt’, ‘bootstrap’: True } Running RandomForestClassifier with suggested hyperparameters: {‘n_estimators’: 100, ‘max_depth’: None, ‘min_samples_split’: 2, ‘min_samples_leaf’: 1, ‘max_features’: ‘sqrt’, ‘bootstrap’: True} Metrics after tuning: {‘accuracy’: 0.9730041532071989, ‘precision’: 0.9722907483489197, ‘recall’: 0.9730041532071989, ‘f1_score’: 0.9724045530119824}<\/p>\n
\u8fd9\u662f\u6211\u8bd5\u9a8c\u5f97\u5230\u7684\u793a\u4f8b\u8f93\u51fa\u3002\u5b83\u53ef\u80fd\u548c\u4f60\u7684\u4e0d\u4e00\u6837\u3002\u4f60\u53ef\u4ee5\u8bbe\u7f6e\u63d0\u793a\u548c\u751f\u6210\u53c2\u6570\uff0c\u4ee5\u83b7\u5f97\u66f4\u52a0\u591a\u53d8\u6216\u4e25\u683c\u7684LLM\u8f93\u51fa\u3002\u7136\u800c\uff0c\u5982\u679c\u4f60\u6b63\u786e\u6784\u5efa\u4e86\u4ee3\u7801\u7684\u7ed3\u6784\uff0c\u53ef\u4ee5\u5c06LLM\u8fd0\u7528\u4e8e\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u81ea\u52a8\u5316\u3002<\/p>\n
LLM\u5df2\u7ecf\u5e94\u7528\u4e8e\u8bb8\u591a\u4f7f\u7528\u573a\u666f\uff0c\u5305\u62ec\u4ee3\u7801\u751f\u6210\u3002\u901a\u8fc7\u8fd0\u7528LLM\uff08\u6bd4\u5982OpenAI GPT\u6a21\u578b\uff09\uff0c\u6211\u4eec\u5c31\u5f88\u5bb9\u6613\u59d4\u6d3eLLM\u5904\u7406\u6a21\u578b\u9009\u62e9\u548c\u8bd5\u9a8c\u8fd9\u9879\u4efb\u52a1\uff0c\u53ea\u8981\u6211\u4eec\u6b63\u786e\u5730\u6784\u5efa\u8f93\u51fa\u7684\u7ed3\u6784\u3002\u5728\u672c\u4f8b\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u6837\u672c\u6570\u636e\u96c6\u5bf9\u6a21\u578b\u8fdb\u884c\u8bd5\u9a8c\uff0c\u8ba9LLM\u9009\u62e9\u548c\u8bd5\u9a8c\u4ee5\u6539\u8fdb\u6a21\u578b\u3002<\/p>\n
\u6587\u7ae0\u6765\u81ea\uff1a51CTO<\/p>\n