データ分析向けのCSVから列をドロップするだけのアプリつくーる（Windows 10、Python 3.10、PyQt5）

はじめに

CSVから列をドロップするだけのアプリつくりまーす

開発環境

• Windows 10 PC
• Python 3.10
• PyQt5

導入

1．ライブラリのインストール

pip install pandas 
pip install seaborn
pip install PyQt5

使用しているライブラリはこんな感じ

requirements.txt

contourpy==1.1.0
cycler==0.11.0
fonttools==4.42.0
kiwisolver==1.4.4
matplotlib==3.7.2
numpy==1.25.2
packaging==23.1
pandas==2.0.3
Pillow==10.0.0
pyparsing==3.0.9
PyQt5==5.15.9
PyQt5-Qt5==5.15.2
PyQt5-sip==12.12.2
python-dateutil==2.8.2
pytz==2023.3
seaborn==0.12.2
six==1.16.0
tzdata==2023.3

pip freeze > requirements.txt

これで書き出せます。

2．コードを作成。ChatGPTに聞きまくれ！

app.py

import sys
import pandas as pd
import seaborn as sns

from PyQt5 import QtWidgets
from PyQt5.QtWidgets import QFileDialog, QTabWidget
from PyQt5.QtCore import Qt
from PyQt5.QtGui import QIcon

from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
from matplotlib.figure import Figure


class MplCanvas(FigureCanvas):
    def __init__(self, parent=None, width=10, height=8, dpi=100):
        fig = Figure(figsize=(width, height), dpi=dpi)
        self.axes = fig.add_subplot(111)
        super(MplCanvas, self).__init__(fig)


class FeatureDropperApp(QtWidgets.QWidget):
    def __init__(self):
        super().__init__()

        # Load Button
        self.load_button = QtWidgets.QPushButton("Load CSV File")
        self.load_button.clicked.connect(self.load_csv)

        # Features List
        self.features_list = QtWidgets.QListWidget()
        self.features_list.setSelectionMode(QtWidgets.QAbstractItemView.MultiSelection)

        # Drop Button
        self.drop_button = QtWidgets.QPushButton("Drop Selected Features")
        self.drop_button.clicked.connect(self.drop_features)

        # Save Button
        self.save_button = QtWidgets.QPushButton("Save As CSV")
        self.save_button.clicked.connect(self.save_csv)

        # Layout for left side
        left_layout = QtWidgets.QVBoxLayout()
        left_layout.addWidget(self.load_button)
        left_layout.addWidget(self.features_list)
        left_layout.addWidget(self.drop_button)
        left_layout.addWidget(self.save_button)

        # Tab Widget (right side)
        self.tab_widget = QTabWidget()
        self.distribution_scroll_area = QtWidgets.QScrollArea()
        self.distribution_scroll_area.setHorizontalScrollBarPolicy(Qt.ScrollBarAsNeeded)
        self.correlation_canvas = MplCanvas(self, width=6, height=6, dpi=100)
        self.missing_values_canvas = MplCanvas(self, width=6, height=6, dpi=100)

        # Add a table for the first 100 rows
        self.first_100_rows_table = QtWidgets.QTableWidget()
        self.first_100_rows_table.setEditTriggers(
            QtWidgets.QAbstractItemView.NoEditTriggers
        )

        # Statistics Tab with Numerical and Categorical sub-tabs
        self.statistics_tab_widget = QTabWidget()
        self.numerical_statistics_layout = QtWidgets.QVBoxLayout()
        self.numerical_statistics_widget = QtWidgets.QWidget()
        self.numerical_statistics_widget.setLayout(self.numerical_statistics_layout)
        self.categorical_statistics_layout = QtWidgets.QVBoxLayout()
        self.categorical_statistics_widget = QtWidgets.QWidget()
        self.categorical_statistics_widget.setLayout(self.categorical_statistics_layout)
        self.statistics_tab_widget.addTab(self.numerical_statistics_widget, "Numerical")
        self.statistics_tab_widget.addTab(
            self.categorical_statistics_widget, "Categorical"
        )

        self.tab_widget.addTab(self.first_100_rows_table, "First 100 Rows")
        self.tab_widget.addTab(self.statistics_tab_widget, "Statistics")
        self.tab_widget.addTab(self.distribution_scroll_area, "Distribution")
        self.tab_widget.addTab(self.correlation_canvas, "Correlation")
        self.tab_widget.addTab(self.missing_values_canvas, "Missing Values")

        # Main layout
        main_layout = QtWidgets.QHBoxLayout()
        main_layout.addLayout(left_layout, 1)
        main_layout.addWidget(self.tab_widget, 3)

        self.setLayout(main_layout)
        self.data = None
        self.resize(800, 600)  # adjust the numbers as you see fit
        self.setWindowTitle("Feature Dropper")
        self.setWindowIcon(QIcon("favicon.ico"))

    def show_first_100_rows(self):
        if self.data is not None:
            # Display the first 100 rows in the table
            self.first_100_rows_table.setRowCount(min(100, len(self.data)))
            self.first_100_rows_table.setColumnCount(len(self.data.columns))
            self.first_100_rows_table.setHorizontalHeaderLabels(self.data.columns)

            for i in range(min(100, len(self.data))):
                for j, value in enumerate(self.data.iloc[i]):
                    self.first_100_rows_table.setItem(
                        i, j, QtWidgets.QTableWidgetItem(str(value))
                    )

            self.first_100_rows_table.resizeColumnsToContents()

    def show_statistics(self):
        if self.data is not None:
            # Get numerical and categorical columns
            numerical_data = self.data.select_dtypes(include=["number"])
            categorical_data = self.data.select_dtypes(include=["object", "category"])

            # Clear the previous contents
            for i in reversed(range(self.numerical_statistics_layout.count())):
                self.numerical_statistics_layout.itemAt(i).widget().setParent(None)
            for i in reversed(range(self.categorical_statistics_layout.count())):
                self.categorical_statistics_layout.itemAt(i).widget().setParent(None)

            # Numerical data
            numerical_table = QtWidgets.QTableWidget()
            numerical_table.setColumnCount(len(numerical_data.columns))
            numerical_table.setRowCount(4)
            numerical_table.setHorizontalHeaderLabels(numerical_data.columns)
            numerical_table.setVerticalHeaderLabels(["Mean", "Median", "Min", "Max"])
            numerical_table.setEditTriggers(QtWidgets.QAbstractItemView.NoEditTriggers)

            for i, col in enumerate(numerical_data.columns):
                numerical_table.setItem(
                    0, i, QtWidgets.QTableWidgetItem(str(numerical_data[col].mean()))
                )
                numerical_table.setItem(
                    1, i, QtWidgets.QTableWidgetItem(str(numerical_data[col].median()))
                )
                numerical_table.setItem(
                    2, i, QtWidgets.QTableWidgetItem(str(numerical_data[col].min()))
                )
                numerical_table.setItem(
                    3, i, QtWidgets.QTableWidgetItem(str(numerical_data[col].max()))
                )

            self.numerical_statistics_layout.addWidget(numerical_table)

            # Categorical data
            categorical_table = QtWidgets.QTableWidget()
            categorical_table.setColumnCount(len(categorical_data.columns))
            categorical_table.setRowCount(2)
            categorical_table.setHorizontalHeaderLabels(categorical_data.columns)
            categorical_table.setVerticalHeaderLabels(
                ["Unique values", "Most common value"]
            )
            categorical_table.setEditTriggers(
                QtWidgets.QAbstractItemView.NoEditTriggers
            )

            for i, col in enumerate(categorical_data.columns):
                categorical_table.setItem(
                    0,
                    i,
                    QtWidgets.QTableWidgetItem(str(categorical_data[col].nunique())),
                )
                categorical_table.setItem(
                    1,
                    i,
                    QtWidgets.QTableWidgetItem(
                        str(categorical_data[col].mode().iat[0])
                    ),
                )

            self.categorical_statistics_layout.addWidget(categorical_table)

    def show_distribution(self):
        if self.data is not None:
            # Get selected features
            selected_features = self.data.columns.tolist()
            numerical_data = self.data[selected_features].select_dtypes(
                include=["number"]
            )

            # Create a widget and a vertical layout to hold the histograms
            distribution_widget = QtWidgets.QWidget()
            distribution_layout = QtWidgets.QHBoxLayout()

            for col in numerical_data.columns:
                # Create a new MplCanvas for each histogram
                canvas = MplCanvas(self, width=5, height=4, dpi=100)
                numerical_data[col].plot(kind="hist", ax=canvas.axes)
                canvas.axes.set_title(col)

                # Rotate x-axis labels
                for label in canvas.axes.get_xticklabels():
                    label.set_rotation(45)

                # Add the canvas to the layout
                distribution_layout.addWidget(canvas)

                # Adjust layout to make sure everything fits
                canvas.figure.tight_layout()

            # Set the layout on the widget and set the widget on the scroll area
            distribution_widget.setLayout(distribution_layout)
            self.distribution_scroll_area.setWidget(distribution_widget)

    def show_correlation(self):
        if self.data is not None:
            self.correlation_canvas.axes.clear()
            numerical_data = self.data.select_dtypes(include=["number"])
            corr = numerical_data.corr()
            sns.heatmap(corr, ax=self.correlation_canvas.axes)
            self.correlation_canvas.figure.tight_layout()
            self.correlation_canvas.draw()

    def show_missing_values(self):
        if self.data is not None:
            self.missing_values_canvas.axes.clear()
            missing_values = self.data.isnull().sum()
            missing_values = missing_values[missing_values > 0]
            missing_values.plot(kind="bar", ax=self.missing_values_canvas.axes)
            self.missing_values_canvas.figure.tight_layout()
            self.missing_values_canvas.draw()

    def load_csv(self):
        options = QFileDialog.Options()
        file_name, _ = QFileDialog.getOpenFileName(
            self,
            "Load CSV File",
            "",
            "CSV Files (*.csv);;All Files (*)",
            options=options,
        )
        if file_name:
            self.data = pd.read_csv(file_name)
            self.features_list.addItems(self.data.columns.tolist())
            self.show_first_100_rows()
            self.show_statistics()
            self.show_distribution()
            self.show_correlation()
            self.show_missing_values()

    def drop_features(self):
        if self.data is not None:
            selected_features = [
                item.text() for item in self.features_list.selectedItems()
            ]
            self.data.drop(columns=selected_features, inplace=True)
            self.features_list.clear()
            self.features_list.addItems(self.data.columns.tolist())
            self.show_first_100_rows()
            self.show_statistics()
            self.show_distribution()
            self.show_correlation()
            self.show_missing_values()

    def save_csv(self):
        if self.data is not None:
            options = QFileDialog.Options()
            file_name, _ = QFileDialog.getSaveFileName(
                self,
                "Save CSV File",
                "",
                "CSV Files (*.csv);;All Files (*)",
                options=options,
            )
            if file_name:
                self.data.to_csv(file_name, index=False)


app = QtWidgets.QApplication(sys.argv)
window = FeatureDropperApp()
window.show()
sys.exit(app.exec_())

実行

1．実行すると以下のように表示されます

python app.py

2．Load CSV FileからCSVファイルを読み込みましょう

CSVファイルはKaggleのAsteroid Datasetを用います。

435MBもあります。Visual Studio CodeのEdit csvだと50MBを超えるファイルは編集できないです。

3．First 100 Rows：最初の100行を表示します

4．Statistics：数値データとカテゴリデータの統計データを表示します

Numerical	Categorical

5．Distribution：データの分布を表示します

6．Correlation：データの相関のグラフを表示します

7．Missing Values：欠損値の個数を表示します

8．欠損値の多いnameとprefixを選択してドロップします

Drop Selected Featuresをクリックするとグラフも更新されます

9．Save As CSVで保存します

EXE化して配布

アプリのアイコン（favicon.ico）を用意します。
アプリ名はFeatureDropperにしました。

pip install pyinstaller
pyinstaller --onefile --noconsole --icon=favicon.ico --name=FeatureDropper app.py

実行するとdistフォルダの中にFeatureDropper.exeが生成されます。
配布して実行してもらうと不明な発行元の警告が出ます。
解決するには証明書が必要みたいです。

お疲れ様でした。
Powered by ChatGPT