Finetune_Gemma_Model / pages /Dataset_Management.py
mohitkumarrajbadi's picture
New Improvement in Pages
e37cfd0
import streamlit as st
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import os
from utils import (
load_dataset,
save_dataset,
clean_dataset,
compute_dataset_score,
detect_outliers,
apply_transformation,
list_datasets,
detect_inconsistent_types
)
# -------------------------------
# Constants & Setup
# -------------------------------
DATASET_DIR = "datasets"
DEFAULT_DATASET = "train_data.csv"
os.makedirs(DATASET_DIR, exist_ok=True) # Ensure directory exists
# -------------------------------
# Sidebar: Dataset Selection
# -------------------------------
st.sidebar.header("πŸ“Š Dataset Selection")
# List available datasets from the datasets folder
available_datasets = list_datasets(DATASET_DIR)
dataset_choice = st.sidebar.radio("Choose Dataset Source:", ["Select Existing Dataset", "Upload New Dataset"])
dataset_path = None
if dataset_choice == "Select Existing Dataset":
if available_datasets:
selected_dataset = st.sidebar.selectbox("Select Dataset:", available_datasets)
dataset_path = os.path.join(DATASET_DIR, selected_dataset)
st.sidebar.success(f"Using `{selected_dataset}` dataset.")
else:
st.sidebar.warning("No datasets found. Please upload a new dataset.")
elif dataset_choice == "Upload New Dataset":
uploaded_file = st.sidebar.file_uploader("Upload Dataset (CSV, JSON, or Excel)", type=["csv", "json", "xlsx"])
if uploaded_file:
file_ext = uploaded_file.name.split('.')[-1].lower()
try:
if file_ext == "csv":
new_df = pd.read_csv(uploaded_file)
elif file_ext == "json":
new_df = pd.json_normalize(json.load(uploaded_file))
elif file_ext == "xlsx":
new_df = pd.read_excel(uploaded_file)
else:
st.error("Unsupported file format.")
st.stop()
except Exception as e:
st.error(f"Error reading file: {e}")
st.stop()
# Save the new dataset with its filename
dataset_path = os.path.join(DATASET_DIR, uploaded_file.name)
save_dataset(new_df, dataset_path)
st.sidebar.success(f"Dataset `{uploaded_file.name}` uploaded successfully!")
available_datasets = list_datasets(DATASET_DIR) # Refresh list
else:
st.sidebar.warning("Please upload a dataset.")
# -------------------------------
# Load the Selected Dataset
# -------------------------------
if dataset_path:
df = load_dataset(dataset_path)
if df.empty:
st.warning("Dataset is empty or failed to load.")
else:
df = pd.DataFrame()
st.warning("No dataset selected. Please choose or upload a dataset.")
# -------------------------------
# Main App Title & Description
# -------------------------------
st.title("πŸ“Š The Data Hub")
# -------------------------------
# Tabs for Operations
# -------------------------------
tabs = st.tabs([
"View & Summary", "Clean Data",
"Visualize Data", "Data Profiling",
"Outlier Detection", "Custom Transformations",
"Export"
])
# -------------------------------
# Tab 1: View & Summary
# -------------------------------
with tabs[0]:
st.subheader("πŸ“‹ Current Dataset Preview")
if not df.empty:
st.dataframe(df)
st.markdown("#### πŸ”Ž Basic Statistics")
st.write(df.describe(include="all"))
else:
st.warning("No dataset available. Please choose or upload a dataset.")
# -------------------------------
# Tab 2: Clean Data
# -------------------------------
with tabs[1]:
st.subheader("🧼 Clean Your Dataset")
if not df.empty:
remove_duplicates = st.checkbox("Remove Duplicate Rows", value=True)
fill_missing = st.checkbox("Fill Missing Values", value=False)
fill_value = st.text_input("Fill missing values with:", value="0")
st.markdown("#### Optional: Rename Columns")
new_names = {}
for col in df.columns:
new_names[col] = st.text_input(f"Rename column '{col}'", value=col)
if st.button("Clean Dataset"):
cleaned_df = clean_dataset(df, remove_duplicates, fill_missing, fill_value)
cleaned_df = cleaned_df.rename(columns=new_names)
save_dataset(cleaned_df, dataset_path)
st.success("βœ… Dataset cleaned successfully!")
st.dataframe(cleaned_df.head())
df = cleaned_df
else:
st.warning("No dataset available for cleaning.")
# -------------------------------
# Tab 3: Visualize Data (Fixed KeyError Issue)
# -------------------------------
with tabs[2]:
st.subheader("πŸ“Š Visualize Your Data")
if not df.empty:
viz_type = st.selectbox("Select Visualization Type", ["Histogram", "Scatter", "Box Plot", "Heatmap", "Line Chart"])
numeric_cols = df.select_dtypes(include=["number"]).columns.tolist()
if numeric_cols:
# Validate column selection
col = st.selectbox("Select Column", numeric_cols)
if col: # Ensure valid column selection
fig, ax = plt.subplots()
if viz_type == "Histogram":
ax.hist(df[col].dropna(), bins=20, color="skyblue", edgecolor="black")
elif viz_type == "Box Plot":
sns.boxplot(x=df[col].dropna(), ax=ax)
elif viz_type == "Scatter":
x_col = st.selectbox("X-axis", numeric_cols)
y_col = st.selectbox("Y-axis", numeric_cols)
if x_col and y_col:
ax.scatter(df[x_col], df[y_col], color="green")
elif viz_type == "Heatmap":
corr = df[numeric_cols].corr()
sns.heatmap(corr, annot=True, cmap="coolwarm", ax=ax)
elif viz_type == "Line Chart":
ax.plot(df.index, df[col], marker="o")
st.pyplot(fig)
else:
st.warning("Please select a valid column.")
else:
st.warning("No numeric columns available for visualization.")
else:
st.warning("No dataset available for visualization.")
# -------------------------------
# Tab 4: Data Profiling
# -------------------------------
with tabs[3]:
if not df.empty:
# -------------------------------
# 1. General Dataset Info
# -------------------------------
st.markdown("### πŸ› οΈ General Information")
st.write(f"βœ… **Total Rows:** `{df.shape[0]}`")
st.write(f"βœ… **Total Columns:** `{df.shape[1]}`")
st.write(f"βœ… **Memory Usage:** `{df.memory_usage(deep=True).sum() / (1024 ** 2):.2f} MB`")
st.write(f"βœ… **Dataset Shape:** `{df.shape}`")
# -------------------------------
# 2. Dataset Quality Score
# -------------------------------
st.markdown("### πŸ“Š Dataset Quality Score")
score = compute_dataset_score(df)
st.success(f"πŸ’― Dataset Quality Score: `{score} / 100`")
# -------------------------------
# 3. Column Overview with Stats
# -------------------------------
st.markdown("### πŸ”₯ Column Overview")
# Numeric and categorical columns
numeric_cols = df.select_dtypes(include=["number"]).columns
categorical_cols = df.select_dtypes(include=["object"]).columns
profile = pd.DataFrame({
"Column": df.columns,
"Data Type": df.dtypes.values,
"Missing Values": df.isnull().sum().values,
"Missing %": (df.isnull().sum() / len(df) * 100).values,
"Unique Values": df.nunique().values
})
# Add numeric statistics
if len(numeric_cols) > 0:
numeric_stats = pd.DataFrame({
"Column": numeric_cols,
"Min": df[numeric_cols].min().values,
"Max": df[numeric_cols].max().values,
"Mean": df[numeric_cols].mean().values,
"Std Dev": df[numeric_cols].std().values,
"Skewness": df[numeric_cols].skew().values,
"Kurtosis": df[numeric_cols].kurt().values
})
# Merge stats with the profile
profile = profile.merge(numeric_stats, on="Column", how="left")
st.dataframe(profile)
# -------------------------------
# 4. Missing Values Visualization
# -------------------------------
st.markdown("### πŸ”Ž Missing Values Distribution")
missing_values = df.isnull().sum()
missing_values = missing_values[missing_values > 0]
if not missing_values.empty:
fig, ax = plt.subplots(figsize=(12, 5))
sns.barplot(x=missing_values.index, y=missing_values.values, ax=ax, color="skyblue")
ax.set_title("Missing Values per Column")
ax.set_ylabel("Missing Count")
ax.set_xticklabels(ax.get_xticklabels(), rotation=45)
st.pyplot(fig)
else:
st.success("No missing values found!")
# -------------------------------
# 5. Duplicates Detection
# -------------------------------
st.markdown("### πŸ”₯ Duplicates & Constant Columns Detection")
# Duplicates
duplicate_count = df.duplicated().sum()
st.write(f"πŸ” **Duplicate Rows:** `{duplicate_count}`")
# Constant Columns
constant_cols = [col for col in df.columns if df[col].nunique() == 1]
if constant_cols:
st.write(f"🚩 **Constant Columns:** `{constant_cols}`")
else:
st.success("No constant columns detected!")
# -------------------------------
# 6. Cardinality Analysis
# -------------------------------
st.markdown("### 🧬 Cardinality Analysis")
high_cardinality = [col for col in df.columns if df[col].nunique() > len(df) * 0.8]
if high_cardinality:
st.write(f"πŸ”’ **High-Cardinality Columns:** `{high_cardinality}`")
else:
st.success("No high-cardinality columns detected!")
# -------------------------------
# 7. Top Frequent & Rare Values
# -------------------------------
st.markdown("### 🎯 Frequent & Rare Values")
for col in categorical_cols:
st.write(f"βœ… **{col}**")
top_values = df[col].value_counts().nlargest(5)
rare_values = df[col].value_counts().nsmallest(5)
st.write("πŸ“Š **Top Frequent Values:**")
st.dataframe(top_values)
st.write("πŸ§ͺ **Rare Values:**")
st.dataframe(rare_values)
# -------------------------------
# 8. Correlation Matrix
# -------------------------------
st.markdown("### πŸ“Š Correlation Matrix")
if len(numeric_cols) > 1:
corr = df[numeric_cols].corr()
fig, ax = plt.subplots(figsize=(12, 8))
sns.heatmap(corr, annot=True, fmt=".2f", cmap="coolwarm", square=True, ax=ax)
st.pyplot(fig)
else:
st.info("Not enough numeric columns for correlation analysis.")
# -------------------------------
# 9. Pair Plot (Numerical Relationships)
# -------------------------------
st.markdown("### πŸ”₯ Pair Plot (Numerical Relationships)")
if len(numeric_cols) >= 2:
pairplot = sns.pairplot(df[numeric_cols], diag_kind='kde')
st.pyplot(pairplot.fig)
else:
st.info("Not enough numeric columns for pair plot visualization.")
# -------------------------------
# 10. Outlier Detection
# -------------------------------
st.markdown("### 🚩 Outlier Detection")
outliers = detect_outliers(df)
if outliers:
st.write("βœ… **Outliers Detected:**")
st.dataframe(pd.DataFrame(outliers.items(), columns=["Column", "Outlier Count"]))
else:
st.success("No significant outliers detected!")
# -------------------------------
# 11. Inconsistent Data Types
# -------------------------------
st.markdown("### 🚫 Inconsistent Data Types")
inconsistent_types = detect_inconsistent_types(df)
if inconsistent_types:
st.write("⚠️ **Inconsistent Data Types Detected:**")
st.write(inconsistent_types)
else:
st.success("No inconsistent data types detected!")
else:
st.warning("No dataset available for profiling.")
# -------------------------------
# Tab 5: Outlier Detection
# -------------------------------
with tabs[4]:
st.subheader("πŸš€ Outlier Detection")
if not df.empty:
outliers = detect_outliers(df)
st.write(outliers)
else:
st.warning("No dataset available for outlier detection.")
# -------------------------------
# Tab 6: Export
# -------------------------------
with tabs[5]:
st.subheader("πŸ“€ Export Dataset")
export_format = st.selectbox("Export Format", ["CSV", "Excel", "JSON"])
if not df.empty:
st.download_button("Download", df.to_csv(index=False), f"dataset.{export_format.lower()}")