このサイトを正しく表示するには、最新のブラウザを使用し、Javascriptを有効化してください。
最高の機械学習

ねこ

import asyncio
import discord
from discord.ext import commands, tasks
from discord import Interaction, Member, app_commands, ui
from discord.app_commands import Group, command
from discord.ext.commands import GroupCog
import traceback
import os
import csv
from sklearn.model_selection import train_test_split
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score

class ac_learn(commands.Cog):
    def __init__(self, bot):
        # 必要なファイルを読み取る
        self.bot = bot
        self.message_id = {
            0: '問題なし',
            1: '嫌がらせ',
            2: '不適切な内容',
            3: '詐欺・スパム'
        }

@commands.Cog.listener()
async def on_ready(self):
    self.check = app_commands.ContextMenu(
        name='check',
        callback=self.check_text
    )
    self.bot.tree.add_command(self.check)

    self.learn = app_commands.ContextMenu(
        name='learn',
        callback=self.learn_text
    )
    self.bot.tree.add_command(self.learn)

async def check_text(self, interaction: discord.Interaction, message: discord.Message) -> None:
    await interaction.response.defer(ephemeral=True)
    try:
        vanguard = Vanguard()
        res = vanguard.check(message.content)
        pred = self.message_id[res[0]]
        embed = discord.Embed(
            title=f"{interaction.user.display_name} によって送信されたメッセージ",
            description=f"{message.content}\n\nこのメッセージの判定は\n{pred}",
            color=0x00ff00
        )
        if res[0] == 0:
            embed.color = 0x00ff00
        else:
            embed.color = 0xff0000
        embed.set_thumbnail(url=message.author.display_avatar)
        await interaction.followup.send(embed=embed)
    except Exception as e:
        embed = discord.Embed(
            title="不明なコマンド",
            description=f"エラーが発生しました。\n{e}",
            color=0xff0000
        )
        await interaction.followup.send(embed=embed)

async def learn_text(self, interaction: discord.Interaction, message: discord.Message) -> None:
    embed = discord.Embed(
        title=f"{interaction.user.display_name} によって送信されたメッセージ",
        description=f"{message.content}",
        color=0x00ff00
    )
    embed.set_thumbnail(url=message.author.display_avatar)
    try:
        view = discord.ui.View(timeout=None)
        view.add_item(SelectList(self.bot, interaction, message))
        await interaction.response.send_message(embed=embed, view=view, ephemeral=True)
    except Exception as e:
        print(e)

class SelectList(discord.ui.Select):
    def __init__(self, bot, interaction, message):
        self.bot = bot
        self.path = '/home/neko/pylearn/suzuka-vanguard.csv'
        self.message = message
        options = []
        options.append(discord.SelectOption(
            label="問題なし",
            value="0",
            description="このチャットテキストは問題ありません"
        ))
        options.append(discord.SelectOption(
            label="嫌がらせ",
            value="1",
            description="なりすましを含む、個人に対する攻撃"
        ))
        options.append(discord.SelectOption(
            label="不適切な内容",
            value="2",
            description="迷惑行為および、不適切な内容、下品な言葉の使用、悪態等"
        ))
        options.append(discord.SelectOption(
            label="詐欺・スパム",
            value="3",
            description="詐欺、スパムなど。刑務所にぶちこまれる楽しみにしておいて下さい！いいですね！"
        ))
        # options.append(discord.SelectOption(
        #     label="",
        #     value="",
        #     description=""
        # ))
        # options.append(discord.SelectOption(
        #     label="",
        #     value="",
        #     description=""
        # ))
        # options.append(discord.SelectOption(
        #     label="",
        #     value="",
        #     description=""
        # ))

    super().__init__(placeholder="このテキストチャットはどうですか？", options=options, min_values=1, max_values=1)

async def callback(self, interaction: discord.Integration):
    await interaction.response.defer(ephemeral=True)
    try:
        if (interaction.user.id in self.bot.owner_ids):
            values = int(self.values[0])
            with open(self.path, 'a', newline='', encoding='utf-8') as csvfile:
                writer = csv.writer(csvfile)
                for text, label in zip([self.message.content], [values]):
                    writer.writerow([text, label])

            embed = discord.Embed(
                title="不明なコマンド",
                description=f"書き込みました",
                color=0x00ff00
            )
            await interaction.followup.send(embed=embed, ephemeral=True)
        else:
            raise("権限がありません")
    except Exception as e:
        embed = discord.Embed(
            title="不明なコマンド",
            description=f"エラーが発生しました。\n{e}",
            color=0xff0000
        )
        await interaction.followup.send(embed=embed, ephemeral=True)

class Vanguard:
    def __init__(self):
        # ロード
        self.texts = []
        self.labels = []
        self.path = '/home/neko/pylearn/suzuka-vanguard.csv'
        with open(self.path, mode='r', newline='', encoding='utf-8') as file:
            reader = csv.reader(file)
            for row in reader:
                self.texts.append(row[0])
                self.labels.append(int(row[1]))

def check(self, message):
    X_train, X_test, y_train, y_test = train_test_split(self.texts, self.labels, test_size=0.2, random_state=42)

    # TF-IDFベクトル化
    vectorizer = TfidfVectorizer()
    X_train_vectorized = vectorizer.fit_transform(X_train)
    X_test_vectorized = vectorizer.transform(X_test)
    # モデルのトレーニング（ロジスティック回帰を使用）
    # model = LogisticRegression()
    # model.fit(X, self.labels)
    svm_classifier = SVC(kernel='linear')
    svm_classifier.fit(X_train_vectorized, y_train)

    # テストデータで予測
    predictions = svm_classifier.predict(X_test_vectorized)

    # 精度の評価
    accuracy = accuracy_score(y_test, predictions)
    print(f"精度: {accuracy}")

    # 分類したい文字列
    new_text = [message]

    # ベクトル化
    # new_text_vectorized = vectorizer.transform([new_text])
    # モデルで分類
    # prediction = model.predict(new_text_vectorized)

    new_message_vectorized = vectorizer.transform(new_text)
    prediction_new = svm_classifier.predict(new_message_vectorized)


    return prediction_new

async def setup(bot):
    await bot.add_cog(ac_learn(bot))