[🇮🇹 Italiano] [🇧🇷 Português] [🇺🇸 English]

⚡️ CDIA Nexus PUC-SP: Innovation Hub for Smart Water and Energy Solutions ⚡️ Smart City Laguna IoT, Fortaleza, Brazil ⚡️

Project for monitoring, forecasting, and optimizing energy consumption in a smart home, using IoT and AI. Developed in the context of Smart City Laguna – CDIA PUC-SP.

In collaboration with Planet Smart City, PUC-SP - Data Science & AI, UN Sustainable Development Goals (SDGs), Starlink and Proptech Brazil

Connected.4.Good.-.IoT.in.Planet.SmartCity.mp4

📺 Watch in Full HD on YouTube

⚠️ Heads Up

• Projects and deliverables may be made publicly available whenever possible.

• The course prioritizes hands-on practice with real data in consulting scenarios.

• All activities comply with the academic and ethical guidelines of PUC-SP.

• Confidential information from this repository remains private in private repositories.

Explore the Simulator and support the Smart Cities AI project

⇩💦

🌐 Project Overview:

Developed by the CDIA group at PUC-SP, this extension project aims to optimize smart resource management systems in Smart City Laguna — combining technology, sustainability, and community innovation to empower underserved regions.

With a strong foundation in interdisciplinary collaboration and international cooperation, this initiative bridges data science with real-world applications to foster resilient, inclusive, and intelligent cities.

Planet Smart City:

Founded in 2015 by Giovanni Savio and Susanna Marchionni, Planet Smart City leads the global movement for affordable, smart, and sustainable housing. Their projects combine:

• Advanced urban design
• Integrated technology
• Community-building initiatives

⚡️ CDIA Nexus PUC-SP: Innovation Hub for Smart Water and Energy

The CDIA Nexus is an initiative by the Data Science and Artificial Intelligence Group at PUC-SP, dedicated to developing applied AI and IoT solutions for smart water and energy management.

This innovation hub integrates applied research, university outreach, and social impact, focusing on transforming communities through purposeful technology.

The solutions developed are applied in real contexts, such as Smart City Laguna (Fortaleza, Brazil), through projects in partnership with organizations like Planet Smart City, UN-Habitat, Starlink, among others.

The project aims to combine sustainability, digital inclusion, and social innovation, promoting cities that are more resilient, efficient, and people-centered.

💥 From Code to Insight: Data Analysis and Decision Support:

Project Goal:

To develop a data science and AI-based solution to monitor, forecast, and optimize electricity consumption in a smart home (Smart City Laguna). The project simulates sensor data per room and uses machine learning to anticipate consumption patterns and propose saving actions.

Dataset:

☞ Tap here to access the dataset!

A simulated dataset was used, containing daily records with the following variables:

• Date: Day of measurement
• KW/H: Total energy consumption in kWh
• Room1, Room2, LivingRoom, Kitchen, Pool: Number of sensor activations per room
• SolarGeneration: Energy generated by solar panels (simulated)

Business Question:

“How can we predict daily energy consumption based on room-specific behavior and, from that, propose automated measures for energy savings and efficiency?”

Methodology and Steps:

1. Data import and visualization

   Reading the spreadsheet using pandas and validating formats.

2. Preprocessing

   • Converting the Date column to datetime format
   • Creating the OrdinalDay variable for modeling
   • Calculating average consumption per activation per room
   • Simulating solar generation and projecting future consumption

3. Predictive Modeling

   A Linear Regression model was trained to estimate consumption (KW/H) based on total room activations. It also includes next-day prediction.

4. Visualizations

   • Time series plots with matplotlib/seaborn
   • Ranking of rooms with highest consumption
   • Activation patterns by cluster
   • Interactive dashboard using Streamlit for real-time monitoring (optional)

5. Report Export

   Automatic generation of PDF reports with relevant data, charts, and forecasts.

Results:

• The regression model showed good ability to predict consumption based on room activity
• Living Room and Kitchen were identified as the highest impact areas
• The Pool, although rarely activated, had high average consumption per activation — indicating waste; it was removed from the model, since the Laguna project focuses on social housing and does not include pools
• Solar generation can significantly offset peak-time consumption if properly managed

Conclusions and Recommendations:

• Automate power shut-offs in high-usage areas like the living room and kitchen to achieve immediate savings
• Schedule pool usage to mitigate unnecessary consumption peaks
• Leverage solar generation to balance appliance usage during peak production hours
• Implement alerts when daily consumption targets are exceeded

Deliverables:

• Streamlit app for real-time sensor monitoring
• PDF report with consumption metrics and recommendations
• Notebook containing full data pipeline, predictive model, and visual analyses

Features:

• Real-time dashboard displaying room-based sensor data
• Daily energy consumption forecasting via Linear Regression
• Simulated sensors per room (Room1, Room2, Living Room, Kitchen)
• Daily consumption target with alert system
• Auto-refresh system using streamlit_autorefresh
• Usage pattern clustering using KMeans + PCA
• PDF report export
• Comparison with simulated solar generation

Technologies Used:

• Python
• Pandas and NumPy – data processing and analysis
• Scikit-learn – linear regression and KMeans
• Matplotlib, Seaborn, and Plotly – visualizations
• Streamlit – interactive dashboard
• FPDF – PDF report generation
• Pillow – dashboard image rendering

Project Structure:

laguna_city_digital/
├── app.py # Main Streamlit app
├── consumo_model.pkl # Trained prediction model
├── cluster_model.pkl # Trained KMeans model
├── dados/
│ └── Consumo_de_Energia_Analise.xlsx # Simulated room data
├── relatorios/
│ └── relatorio_consumo_YYYY-MM-DD.pdf
├── imagens/
│ ├── grafico_pca.png
│ ├── heatmap_cluster.png
│ └── grafico_regressao.png
└── README.md

Visualization Examples:

• PCA Clustering
  Distribution of usage patterns by energy profile

• Activation Heatmap
  Usage percentage by room

• Actual vs Predicted Chart
  Evaluation of prediction accuracy

Model Performance:

• R²: 0.70
• RMSE: 11,528.06
• Most influential room: Living Room (28.21%)

Final Conclusions:

Personalized consumption targets

Real-time alert system

• Support for urban energy sustainability
• Scalable foundation for full Smart City deployment

📌 This analysis was developed using data science practices applied to residential energy consumption, aiming to support decision-making for the end user.

Presentation Overview:

CDIA Nexus is the final academic and social extension project of the PUC-SP Data Science and Artificial Intelligence Group, focused on applying IoT and AI for smart water and energy systems in Smart City Laguna, a pioneering urban development in Fortaleza, Brazil.

This initiative was developed in partnership with Planet Smart City, UN-Habitat, and Starlink, aligned with the UN Sustainable Development Goals (SDGs) and committed to social innovation, digital inclusion, and environmental intelligence.

The presentation highlights:

• An integrated Water & Energy Monitoring Dashboard
• Predictive analytics with AI models
• Community engagement through data-driven strategies
• Deployment insights using Starlink connectivity and Planet infrastructure

“Data 4 Good. Innovation with Meaning.”

🌟 Key contributor: Stefano Buono, physicist and entrepreneur, former CERN researcher and founder of AAA (sold to Novartis), now President of LIFTT and CEO of Newcleo (clean nuclear innovation).

➢ Visit Planet Smart City - Oficial 🇮🇹

➢ Visit Planet Smart City - Brazil 🇧🇷

➢ Visit Planet Smart City - India 🇮🇳

The Laguna Project: Smart Social Innovation:

Located in São Gonçalo do Amarante, Ceará, Fortaleza, Brazil; Smart City Laguna is a Planet’s flagship smart city in Brazil, featuring over 60 smart solutions, including:

• Public Wi-Fi and IoT backbone
• Sustainable urban mobility and lighting
• Rainwater drainage with permeable pavements
• Cultural, educational, and governance programs

At the center of this ecosystem:

The Community Manager — a trained professional dedicated to:

• Mobilizing participatory governance
• Promoting workshops, education, and engagement
• Nurturing social cohesion and long-term stewardship

🌎 Global Partnerships:

Special thanks to Pedro Braida Neto, CEO of Proptech Brazil, for leading with empathy, respect, and integrity. The way you show up for others truly makes all the difference.

Ping Pedro 📲

We extend our heartfelt gratitude to the organizations and individuals who made the implementation of the CDIA PUC-SP possible. Special thanks to:

Organization	Contribution
United Nations (UN)	Funding for the acquisition of solar panels
PUC-SP (CDIA)	IoT and AI design and implementation
UN-Habitat	Technical support and ethical frameworks
Starlink	Satellite internet infrastructure
Planet Smart City	Urban development and on-site support
Proptech Brazil	Local deployment and strategic backing

We also extend our gratitude to:

• Local leaders and community members for their trust and ongoing collaboration.
• The multidisciplinary technical team for their dedication to innovative and sustainable solutions.
• Everyone who contributed, directly or indirectly, to bringing this vision to life.

Together, these partners embody an integrated approach to achieving the Sustainable Development Goals, particularly in emerging regions. 💙🌎

⚡ CDIA PUC-São Paulo Module: Water & Energy Systems 💦

The Water & Energy Module designed by CDIA focuses on the use of IoT and AI for resource optimization. Key features:

• Smart sensors for consumption monitoring
• AI-driven dashboards with predictive alerts
• Visualizations for community awareness
• Scalable resource management models

🧑🏼‍🚀 Team Members:

Name	Role
Andson Ribeiro	Github - Contact
Fabiana 🧬 Campanari	Github - Contact Hub
Leonardo X Fernandes	Github - Contact
Pedro Vyctor Almeida	Github - Contact

💙 All members contributed collaboratively across technical and creative areas. Fabiana 🧬 Campanari also led the project’s identity and visual language.

Innovations & Activities:

• Installation of IoT sensors for water and energy monitoring
• Development of predictive dashboards and alert systems
• Co-creation of a data visualization interface for residents
• Deployment of solar energy pilots supported by UN-Habitat
• Real-time analytics for resource planning and sustainability

Learning Outcomes:

The team gained hands-on experience in:

• Design thinking + participatory methodologies
• Field research in urban infrastructure
• Machine learning and data modeling
• Prototyping and system integration
• Delivering solutions that reflect real-world community needs

Visual Documentation:

📷 Photo Gallery

• drone_view_laguna_2025.jpg – Aerial view of the city
• team_workshop_on_site.jpeg – Field activities with residents
• solar_panels_community.jpeg – UN-backed solar deployment
• iot_dashboard_mockup.png – Dashboard design preview

Presentations

• CDIA_Final_Pitch.pdf – Core insights and results
• UN_SolarInvestment_Laguna.pptx – Stakeholder presentation
• IoT_Architecture_Prototype.pptx – Sensor and data flow architecture

Acknowledgment:

We extend our sincere thanks to Pedro from Proptech, whose guidance and expertise were instrumental throughout this project. His support and vision were essential pillars of our development journey.

Dataset Used

☞ tap here to get the dataset

A simulated dataset was used, containing daily records with the following variables:

• Data: Measurement day
• KW/H: Total energy consumption in kWh
• Quarto1, Quarto2, Sala, Cozinha, Piscina: Number of sensor activations in each room
• Geração Solar: Energy generated by solar panels (simulated)

Predictive Modeling:

A [Linear Regression] model was trained to estimate consumption (KW/H) based on total activations per room. Next-day prediction was also implemented.

Visualizations:
- Time series charts with matplotlib/seaborn.
- Ranking of highest-consuming rooms.
- Representations of activations by cluster.
- Interactive Streamlit dashboard for real-time visualization (optional).

📓 Code Pipeline:

Cell 1 - Import libraries

import locale
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA
from sklearn.cluster import KMeans
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_absolute_error, r2_score, mean_squared_error

Cell 2 - Data loading

# Change the path according to your environment
file_path = "/Users/fabicampanari/Desktop/Project Planet Smart City Laguna/2-CRISP-DM - Project Smart City Laguna/🇧🇷 CRISP-DM_Projeto_Smart_City_Laguna/Consumo_de_Energia_Analise.xlsx"
xls = pd.ExcelFile(file_path)
sheet_names = xls.sheet_names
print(sheet_names)
df = xls.parse('Sheet1')
print(df.head())
df.info()

Cell 3 - Date preprocessing

meses_pt = {
    'jan': '01', 'fev': '02', 'mar': '03', 'abr': '04',
    'mai': '05', 'jun': '06', 'jul': '07',
}
df['Data'] = df['Data'].astype(str)
df['Data'] = df['Data'].str.lower().replace(meses_pt, regex=True)
df['Data'] = pd.to_datetime(df['Data'] + '/2025', format='%d/%m/%Y')

Cell 4 - Descriptive statistics and correlation

summary = df.describe()
correlation = df.corr(numeric_only=True)
print(summary)
print(correlation)

Cell 5: PLOT 1 - Variable distributions

fig, axes = plt.subplots(2, 3, figsize=(15, 10))
axes = axes.flatten()
cols = df.columns[1:7]
for i, col in enumerate(cols):
    sns.histplot(df[col], kde=True, ax=axes[i], bins=10)
    axes[i].set_title(f'Distribution - {col}')
    axes[i].set_xlabel(col)
plt.tight_layout()
plt.suptitle("Variable Distributions", fontsize=16, y=1.02)
plt.show()

(                      Data  Total Consumption (kWh)   Bedroom 1   Bedroom 2  \
 count                  211               211.000000  211.000000  211.000000   
 mean   2025-04-16 00:00:00              1188.317536    9.687204    9.549763   
 min    2025-01-01 00:00:00               644.000000    2.000000    0.000000   
 25%    2025-02-22 12:00:00              1057.000000    8.000000    7.000000   
 50%    2025-04-16 00:00:00              1176.000000    9.000000    9.000000   
 75%    2025-06-07 12:00:00              1324.000000   12.000000   11.500000   
 max    2025-07-30 00:00:00              1667.000000   21.000000   21.000000   
 std                    NaN               197.439318    3.176817    3.073874   
 
        Living Room     Kitchen        Pool  
 count   211.000000  211.000000  211.000000  
 mean      9.445498    9.322275    9.383886  
 min       2.000000    3.000000    1.000000  
 25%       7.000000    7.000000    7.000000  
 50%       9.000000    9.000000    9.000000  
 75%      12.000000   11.000000   11.000000  
 max      17.000000   18.000000   19.000000  
 std       3.033247    2.969757    3.436433  ,
                          Total Consumption (kWh)  Bedroom 1  Bedroom 2  \
 Total Consumption (kWh)                 1.000000   0.521439   0.418033   
 Bedroom 1                               0.521439   1.000000   0.060606   
 Bedroom 2                               0.418033   0.060606   1.000000   
 Living Room                             0.548475   0.169207   0.038469   
 Kitchen                                 0.409667   0.071809   0.049356   
...
 Bedroom 1                   0.169207  0.071809  0.064704  
 Bedroom 2                   0.038469  0.049356  0.012383  
 Living Room                 1.000000  0.136760  0.042904  
 Kitchen                     0.136760  1.000000 -0.083571  
 Pool                        0.042904 -0.083571  1.000000  )

Cell 6: PLOT 2 - Total consumption over time

plt.figure(figsize=(14, 6))
plt.plot(df['Data'], df['KW/H'], label='Total Consumption (KW/H)', color='blue', linewidth=2)
plt.title('Total Energy Consumption Over Time')
plt.xlabel('Date')
plt.ylabel('KW/H')
plt.grid(True)
plt.legend()
plt.tight_layout()
plt.show()

Cell 7: Weekly grouping and PLOT 3 - Weekly activations per room

df['Semana'] = df['Data'].dt.to_period('W').apply(lambda r: r.start_time)
df_semana = df.groupby('Semana')[['Quarto1', 'Quarto2', 'Sala', 'Cozinha', 'Piscina']].sum()
df_semana.plot(figsize=(12, 6), marker='o')
plt.title('Weekly Activations per Room')
plt.ylabel('Number of Activations')
plt.xlabel('Week')
plt.xticks(rotation=45)
plt.grid(True)
plt.tight_layout()
plt.show()

Cell 8: PLOT 4 - Correlation between activations and consumption

correlations = df[['KW/H', 'Quarto1', 'Quarto2', 'Sala', 'Cozinha', 'Piscina']].corr()['KW/H'][1:]
plt.figure(figsize=(10, 5))
sns.barplot(x=correlations.index, y=correlations.values, palette='Oranges_r')
plt.title('Correlation between Activations and Energy Consumption (kWh)')
plt.ylabel('Correlation')
plt.xlabel('Room')
plt.tight_layout()
plt.show()

Cell 9: Predictive modeling - Linear Regression and Evaluation

X = df[['Quarto1', 'Quarto2', 'Sala', 'Cozinha']]
y = df['KW/H']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
model = LinearRegression()
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
r2 = r2_score(y_test, y_pred)
print("Mean Squared Error (MSE):", round(mse, 2))
print("R² Score:", round(r2, 2))

Mean Squared Error (MSE): 11528.06
Coefficient of Determination (R²): 0.7

Cell 10: PLOT 5 - Actual vs Predicted Consumption

plt.figure(figsize=(10, 5))
plt.scatter(y_test, y_pred, alpha=0.7)
plt.plot([y.min(), y.max()], [y.min(), y.max()], 'r--')
plt.xlabel("Actual Consumption (kWh)")
plt.ylabel("Predicted Consumption (kWh)")
plt.title("Actual vs Predicted Consumption")
plt.grid(True)
plt.tight_layout()
plt.show()

Contribution of each room to the prediction (coefficients):
Sala       28.214005
Quarto2    24.141777
Quarto1    23.279552
Cozinha    20.993133
dtype: float64

Cell 11 - Model coefficients

coefficients = pd.Series(model.coef_, index=X.columns)
print("\nContribution of each room to the prediction (coefficients):")
print(coefficients.sort_values(ascending=False))

Shows the weight of each room in the consumption prediction.

Cell 12 - Calculate activation percentages per room

df['Total_activations'] = df[['Quarto1', 'Quarto2', 'Sala', 'Cozinha']].sum(axis=1)
for room in ['Quarto1', 'Quarto2', 'Sala', 'Cozinha', 'Piscina']:
    df[f'{room}_pct'] = df[room] / df['Total_activations']

Cell 13: PLOT 6 - Elbow Method for KMeans

scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)
inertia = []
for k in range(1, 10):
    km = KMeans(n_clusters=k, random_state=42)
    km.fit(X_scaled)
    inertia.append(km.inertia_)
plt.figure(figsize=(8,5))
plt.plot(range(1, 10), inertia, marker='o')
plt.title('Elbow Method')
plt.xlabel('Number of clusters')
plt.ylabel('Inertia')
plt.grid(True)
plt.show()

Cell 14: KMeans and PLOT 7 - Pairplot of clusters

# Apply KMeans with the chosen number of clusters
kmeans = KMeans(n_clusters=3, random_state=42)
df['Cluster'] = kmeans.fit_predict(X_scaled)
sns.pairplot(df, hue='Cluster', vars=['Quarto1', 'Quarto2', 'Sala', 'Cozinha'], palette='tab10')
plt.suptitle("Usage Patterns Grouped by Cluster", y=1.02)
plt.show()

Cell 15 - Average profile per cluster and naming

# Calculate average profile per cluster (with activations and percentages)
col_pcts = [f'{c}_pct' for c in ['Quarto1', 'Quarto2', 'Sala', 'Cozinha']]
perfil_clusters = df.groupby('Cluster')[['Quarto1', 'Quarto2', 'Sala', 'Cozinha', 'KW/H'] + col_pcts].mean()

# Function to name the cluster profile considering consumption and activation percentage
def name_cluster(row):
    mean_kw = df['KW/H'].mean()
    if row['KW/H'] < mean_kw * 0.75:
        total_consumption = '🔵 Low Consumption'
    elif row['KW/H'] > mean_kw * 1.25:
        total_consumption = '🔴 High Consumption'
    else:
        total_consumption = '🟡 Balanced Consumption'
    high = []
    for room in ['Quarto1', 'Quarto2', 'Sala', 'Cozinha']:
        mean_pct = df[f'{room}_pct'].mean()
        if row[f'{room}_pct'] > mean_pct * 1.2:
            high.append(room)
    if total_consumption == '🔵 Low Consumption':
        return total_consumption
    if total_consumption == '🟡 Balanced Consumption':
        if len(high) == 0:
            return total_consumption
        else:
            return f"🟠 High Consumption in {', '.join(high)}"
    if total_consumption == '🔴 High Consumption':
        if len(high) == 0:
            return total_consumption
        else:
            return f"🔴 High Consumption (In {', '.join(high)})"

perfil_clusters['Profile'] = perfil_clusters.apply(name_cluster, axis=1)

Cell 16 - Recommendations dictionary and display by cluster

# Function to map profile to recommendation dictionary key
def map_profile_to_key(profile):
    if profile == '🔵 Low Consumption':
        return profile
    if profile == '🟡 Balanced Consumption':
        return profile
    if profile.startswith('🟠 High Consumption'):
        return '🟠 High Consumption'
    if profile.startswith('🔴 High Consumption'):
        if 'In' in profile:
            idx = profile.index('In') + 3
            text = profile[idx:]
            main = text.split(',')[^0].strip()
            if main in ['Sala']:
                return '🔴 High Consumption (Sala/Cozinha)'
            elif main == 'Cozinha':
                return '🔴 High Consumption (Cozinha)'
            else:
                return '🔴 High Consumption'
        else:
            return '🔴 High Consumption'
    return profile

# Dictionary with recommendations per profile
recommendations = {
    '🔵 Low Consumption': [
        "✅ Maintain current good practices.",
        "🎁 Offer rewards or discounts (gamification).",
        "🔋 Encourage use of solar energy / microgeneration."
    ],
    '🟡 Balanced Consumption': [
        "🔌 Automate device shutdown at fixed times.",
        "🕵️ Install presence sensors in bedrooms and living room.",
        "📊 Send weekly comparative usage reports."
    ],
    '🟠 High Consumption': [
        "🛏️ Automate lights and electronics in high-consumption rooms.",
        "🕵️ Install specific presence sensors for the rooms.",
        "📊 Monitor usage to identify unnecessary peaks."
    ],
    '🔴 High Consumption (Sala/Cozinha)': [
        "💧 Schedule kitchen pump operation outside peak hours.",
        "💡 Encourage conscious use of lighting and electronics.",
        "🧠 Suggest automation and white tariff subscription."
    ],
    '🔴 High Consumption (Cozinha)': [
        "🍳 Check kitchen appliances for excessive consumption.",
        "⏰ Control oven and refrigerator usage times.",
        "💡 Encourage efficient lighting use."
    ]
}

# Display profiles and recommendations
for cluster_id, row in perfil_clusters.iterrows():
    print(f"\n=== Cluster {cluster_id} - {row['Profile']} ===")
    print("📊 Average consumption profile (activations and kWh):")
    print(row[['Quarto1', 'Quarto2', 'Sala', 'Cozinha', 'KW/H']])
    print("\n📈 Average percentage of activations per room (%):")
    print((row[col_pcts] * 100).round(2))
    print("\n💡 Recommendations:")
    key = map_profile_to_key(row['Profile'])
    if key in recommendations:
        for rec in recommendations[key]:
            print("-", rec)
    else:
        print("- No specific recommendations for this profile.")

=== Cluster 0 - 🟡 Balanced Consumption ===
📊 Average consumption profile (activations and kWh):
Quarto1       7.747126
Quarto2       8.034483
Sala          7.908046
Cozinha       8.011494
KW/H       1047.402299
Name: 0, dtype: object

📈 Average percentage of activations per room (%):
Quarto1_pct    24.481434
Quarto2_pct    25.376404
Sala_pct       24.970192
Cozinha_pct     25.17197
Name: 0, dtype: object

💡 Recommendations:
- 🔌 Automate turning off equipment at fixed times.
- 🕵️ Install presence sensors in bedrooms and living room.
- 📊 Send weekly comparative usage reports.

=== Cluster 1 - 🟡 Balanced Consumption ===
📊 Average consumption profile (activations and kWh):
Quarto1       9.830508
...
💡 Recommendations:
- 🔌 Automate turning off equipment at fixed times.
- 🕵️ Install presence sensors in bedrooms and living room.
- 📊 Send weekly comparative usage reports.
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...

Cell 17: PLOT 8 - Boxplot of consumption by cluster

plt.figure(figsize=(7,5))
sns.boxplot(x='Cluster', y='KW/H', data=df)
plt.title('Consumption Distribution (KW/H) by Cluster')
plt.show()

Cell 18: PLOT 9 - Heatmap of percentages by cluster

heatmap_data = perfil_clusters[col_pcts] * 100
plt.figure(figsize=(8, 5))
sns.heatmap(heatmap_data, annot=True, cmap='YlGnBu', fmt=".2f")
plt.title('Percentage of Activations per Room (%)')
plt.xlabel('Rooms')
plt.ylabel('Cluster')
plt.show()

Cell 19: PLOT 10 - Radar chart of rooms by cluster

categories = ['Quarto1', 'Quarto2', 'Sala', 'Cozinha']
angles = np.linspace(0, 2 * np.pi, len(categories), endpoint=False).tolist()
angles += angles[:1]
plt.figure(figsize=(10, 8))
for i, row in perfil_clusters.iterrows():
    values = [row[cat] for cat in categories]
    values += values[:1]
    plt.polar(angles, values, label=f'Cluster {i}')
plt.xticks(angles[:-1], categories)
plt.title('Radar of Rooms by Cluster')
plt.legend()
plt.show()

Cell 20: PLOT 11 - Cluster visualization with PCA

PCA Usage

We applied PCA demonstratively, even with only two clusters, to show how it works in dimensionality reduction and highlighting important variables.

Although not essential here, PCA is useful for datasets with many columns or more than two clusters, improving performance and data visualization.

pca = PCA(n_components=2)
X_pca = pca.fit_transform(X_scaled)
df_plot = pd.DataFrame(X_pca, columns=['Component 1', 'Component 2'])
df_plot['Cluster'] = df['Cluster']
plt.figure(figsize=(8,6))
for cluster in df_plot['Cluster'].unique():
    plt.scatter(
        df_plot[df_plot['Cluster'] == cluster]['Component 1'],
        df_plot[df_plot['Cluster'] == cluster]['Component 2'],
        label=f'Cluster {cluster}'
    )
plt.title('Cluster Visualization with PCA')
plt.xlabel('Component 1')
plt.ylabel('Component 2')
plt.legend()
plt.grid(True)
plt.show()

📊 Interpretation of Graphs and Profiles:

• Variable Distribution: Shows how activations and consumption are distributed.
• Temporal Evolution: Allows identification of consumption trends over the days.
• Weekly Activations: Helps visualize patterns by room.
• Correlation: Shows the strength of the relationship between activations and consumption.
• Actual vs Predicted Consumption: Evaluates the quality of the predictive model.
• Clustering: Identifies groups with similar behavior for personalized recommendations.

💡 Recommendations by Profile:

Profile	Main Recommendations
🔵 Low Consumption	Maintain good practices, encourage solar energy, rewards/gamification
🟡 Balanced Consumption	Automate shutdowns, install presence sensors, comparative reports
🟠 High Consumption	Automate lights/electronics, specific presence sensors, monitor peaks
🔴 High Consumption (Living Room/Kitchen)	Schedule pump outside peak, conscious lighting use, suggest automation and time-of-use tariff
🔴 High Consumption (Kitchen)	Check equipment, control usage times, encourage lighting efficiency

🧭 Conclusion:

The project enables identifying consumption patterns, forecasting future use, and recommending actions for greater energy efficiency, customizing recommendations according to each residence’s usage profile.

Note:
Adjust the Excel file path (file_path) according to your environment.

This analysis was prepared based on data science practices applied to residential energy consumption contexts and aims to facilitate decision-making for the end client.

💌 Let the data flow... Ping Us

• 👩🏻‍🚀 Fabiana Campanari - Shoot me an email

• 🧑🏼‍🚀 PedroVyctor - Hit me up by email

• 👨🏽‍🚀 Andson Ribeiro - Slide into my inbox

🛸๋ My Contacts Hub

────────────── ⊹🔭๋ ──────────────

➣➢➤ Back to Top

Copyright 2024 Mindful-AI-Assistants. Code released under the MIT license.