https://github.com/lucashomuniz/project-11
Data Analysis and Machine Learning Techniques for Liver Disease Prediction
https://github.com/lucashomuniz/project-11
classification-model data-analysis decision-tree healthcare-application knn-algorithm liver-disease-prediction logistic-regression machine-learning python-language python-script random-forest supervised-learning svm-model
Last synced: 3 months ago
JSON representation
Data Analysis and Machine Learning Techniques for Liver Disease Prediction
- Host: GitHub
- URL: https://github.com/lucashomuniz/project-11
- Owner: lucashomuniz
- License: apache-2.0
- Created: 2023-07-10T11:37:49.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2025-01-28T11:54:23.000Z (3 months ago)
- Last Synced: 2025-01-28T12:36:42.305Z (3 months ago)
- Topics: classification-model, data-analysis, decision-tree, healthcare-application, knn-algorithm, liver-disease-prediction, logistic-regression, machine-learning, python-language, python-script, random-forest, supervised-learning, svm-model
- Language: Python
- Homepage:
- Size: 58.6 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# ✅ PROJECT-11
In this project, the goal is to deliver strategic insights to address key business questions for a supermarket undergoing **modernization**. Leveraging data from a newly implemented storage tool, the analysis focuses on transforming raw information into actionable insights to guide **investment decisions**, **resource optimization**, and **customer service enhancement**.
The choice of **SQL** and **Power BI** as core tools ensures a streamlined and efficient analytical process. **SQL**, known for its robustness and scalability in handling large datasets, is utilized to **manipulate**, **organize**, and **structure** data in a logical and reliable manner. Its capabilities in integrating data from multiple tables and resolving inconsistencies establish a strong foundation for analysis. **Power BI** complements this by enabling the visualization and presentation of data insights, ensuring clarity and accessibility for decision-making.
The project is centered around answering critical business questions, including identifying **profitable product lines**, analyzing **customer behavior patterns**, and optimizing **human resources** in operations. This **question-driven approach** ensures data usage is targeted and impactful, generating tangible value for the organization. The integration of **SQL** and **Power BI** creates a reliable, efficient workflow, delivering high-quality insights to support strategic decision-making.
**Keywords**: SQL, PowerBI, PowerQuery, DAX, Google Cloud Platform, Business Analytics, BigQuery, Data Visualization, Data Analysis.
# ✅ PROCESS
In **Phase 1**, **data transformations** and **manipulations** are conducted using **SQL** within the **Google Cloud Platform (GCP)** environment. The process begins by addressing the **date inconsistencies** in the “Fato_Vendas” table, where all dates were incorrectly recorded as 2019 due to a system issue. These dates are updated to **2022**, retaining their original days and months. A new column, **“vlr_margem_bruta”**, is then created, calculated from **“vlr_preco_total”**, which incorporates unit price and tax. To ensure accuracy, null values in the **“vlr_preco_unitario”** column are corrected using the **average price of the last 20 days**, segmented by **gender**, **product line**, and **branch**. Following these corrections, the **gross margin** calculation is completed. The phase concludes with the creation of a table to distribute a **monthly investment** of **R$ 10,000 per branch**, based on the **number of products sold per product line and month**. This table includes details on **relative participation** and the exact **investment allocation**.
**Phase 2** focuses on **visualizations** and **analyses** in **Power BI**, leveraging direct connections to the GCP-treated tables. The initial objective is to identify the **most promising product line** for **investment in 2023**, evaluating factors such as **monthly sales growth**, **consumer gender**, and **overall performance**. Additionally, **peak demand periods** are analyzed by examining **hourly sales patterns**, enabling the optimal allocation of **outsourced employees** during critical hours. Lastly, **customer satisfaction** by branch is assessed to uncover **service pattern variations** and identify characteristics contributing to potential issues.
# ✅ CONCLUSION
**Phase 1**: The datasets **“fato_vendas”**, **“dim_produto”**, and **“dim_margem_bruta_perc”** were uploaded to the **Google Cloud Platform (GCP)**, enabling data processing in **BigQuery**. This high-performance, managed environment facilitated handling large datasets, reduced infrastructure maintenance complexity, and prepared for future integrations, such as direct **Power BI** connections. Tasks such as **data cleaning**, **standardization**, and **key metric calculations** established a reliable foundation for advanced analytics. BigQuery's distributed architecture ensured fast query execution, while centralizing workflows in GCP eliminated manual pipelines, accelerating **dashboard development** and updates for enhanced productivity.
**Phase 1A**: **Common Table Expressions (CTEs)** were used to modularize database adjustments, improving organization and readability. The focus was on correcting sales dates, updating unit prices, and creating a new **gross margin** column. In **STEP A**, incorrectly recorded dates in **2019** were updated to **2022** using the CTE **“base_data”**, which applied **DATE_ADD** to shift the year forward by three.
**Phase 1B**: In **STEP B1**, null values in **“vlr_preco_unitario”** were corrected by calculating the **average price** from the past 20 days, segmented by **gender**, **product line**, and **branch**, using the CTE **“average_prices.”** These results were applied in the CTE **“corrected_prices”**, filling most missing values, with exceptions where insufficient historical data was available. In **STEP B2**, the **“vlr_margem_bruta”** column was created by combining **tax values** from **dim_produtos** with **gross margin percentages** from **dim_margem_bruta_perc.** This completed the adjustments to dates, prices, and margin calculations.
**Phase 1C**: A monthly investment of **R$10,000 per branch** was distributed based on the number of products sold by **product line** and **month.** The CTE **“vendas_resumidas”** calculated product quantities sold per branch, while **“produtos_totais”** determined total sales per branch and month, enabling relative participation calculations in **“participacao_produtos.”** The CTE **“investimentos_calculados”** proportionally distributed investments, and **“final_table”** joined with **dim_produtos** to display product names alongside their allocated investments.
**Phase 2**: A **Power BI dashboard** was developed to visualize results from BigQuery. With most transformations handled in **GCP**, Power Query overhead was reduced, streamlining model adjustments and enabling near-real-time updates. The dashboard featured dedicated tabs for each business question, allowing users to explore **segmented metrics** and visuals.
**Phase 2A**: A **DAX measure**, **“m_faturamento”**, calculated revenue by multiplying quantities sold by unit prices using **SUMX.** A parameter allowed users to select the primary metric (**Revenue**, **Unit Price**, or **Product Quantity**) for each visualization. Analysis revealed significant growth for **PRODUCT 6** among female customers, rising from **R$57,000 in January** to **R$152,000 in March**, and for **PRODUCT 3** among male customers, increasing from **R$42,000** to **R$151,000** in the same period. These insights highlighted the strategic importance of these products for maximizing financial returns across gender segments.
**Phase 2B**: A calculated column, **“periodo_venda”**, was created to group sales into **one-hour intervals.** Using the same **parameter logic** as in previous analyses, the primary metric can be filtered dynamically. Revenue analysis identified the **7:00 PM to 8:00 PM** time slot as the most profitable, generating **R$342,170** across all branches. This period represents the **peak hour** for **Branch B** and is also one of the highest-performing periods for **Branch A.** For **Branch A**, the **12:00 PM to 1:00 PM** interval stands out as a key revenue driver, while **Branch C** experiences its highest profitability between **10:00 AM and 11:00 AM.** These insights inform the optimal allocation of **outsourced staff** to address high-demand periods, prioritizing **7:00 PM to 8:00 PM** for most branches, while also emphasizing **12:00 PM to 1:00 PM** and **10:00 AM to 11:00 AM** for specific locations.
**Phase 2C**: This analysis evaluated variables such as **“num_avaliacao,”** **“nom_filial,”** and related characteristics to detect potential satisfaction issues. The findings reveal that all branches exhibit relatively **similar score distributions**, with **Branch C** standing out for having the **highest number of 10-rated evaluations** and the **lowest incidence of 2-rated scores.** While some differences were observed, such as the predominance of **"member"** or **"normal"** clients and variations in the **most evaluated products**, no significant dissatisfaction trends were identified. **Branch C** demonstrates **stronger engagement** and **positive evaluation outcomes**, suggesting an opportunity to share its **best practices** with other branches rather than addressing isolated shortcomings.
