Clustering and Time Series Analysis Project

This project is divided into two main analyses:

1. Clustering Analysis: A project analyzing land mines data using various clustering techniques.
2. Time Series Analysis: A project focused on forecasting oil sales using time series models.

Each section provides in-depth analysis and insights into real-world datasets, using R for statistical computing and visualization.

1. Clustering Analysis

Overview

This analysis applies unsupervised learning techniques to the Land Mines Dataset to identify natural groupings or clusters. The dataset consists of features representing different attributes of land mines, and the goal is to discover distinct clusters based on these attributes.

Methodology

• Data Preprocessing: The dataset was cleaned and preprocessed, with features standardized to ensure meaningful distance metrics.
• Clustering Techniques: Several algorithms were applied to group the data:
  • K-Means Clustering: Identifies k distinct clusters by minimizing the within-cluster variance.
  • Hierarchical Clustering: Builds a hierarchy of clusters using an agglomerative approach, visualized via dendrograms.
  • DBSCAN: Density-based clustering method used to find clusters of varying shapes.
• Evaluation: The optimal number of clusters was chosen using the Elbow Method and Silhouette Scores.

Visualizations:

The following plots are used to visualize the clustering results:

Figure 1: Visualization of the clusters formed using K-Means.

Figure 2: Elbow plot showing the Within-Cluster Sum of Squares (WCSS) vs the number of clusters. This helps determine the optimal number of clusters.

Findings

The analysis revealed distinct groups within the dataset, each corresponding to different types of land mines based on their features.

Files

• Project 1 - Rmd.Rmd: Contains the R code and methodology used for the clustering analysis.
• Project-1---Rmd.html: An HTML report generated from the R markdown file, detailing the analysis.
• Project 1 - Report.pdf: A PDF summary of the clustering project, including key results and visualizations.
• mine_data.xlsx: The dataset used in the clustering analysis.

2. Time Series Analysis

Overview

The goal of this project was to analyze and forecast oil sales using time series data. The dataset contained historical oil sales data, and the objective was to build predictive models to forecast future sales.

Methodology

• Data Exploration: Key patterns in the dataset were explored, including trend, seasonality, and autocorrelation. ACF and PACF plots were used to understand lag relationships.

  

  Figure 3: Plot of the original oil sales data, highlighting missing values.

• Data Imputation: Missing values were handled using imputation strategies. Below is a plot of the data after imputation.

  

  Figure 4: Data plot after imputation using the "up" strategy in R.

• Modeling: Several time series models were implemented:

  • ARIMA (AutoRegressive Integrated Moving Average): A statistical model used for forecasting time series data by combining autoregressive, moving average, and differencing techniques.
  • Exponential Smoothing (ETS): Captures level, trend, and seasonal components in the data.

  

  Figure 5: Explanation of the ETS and Holt-Winters methods used for forecasting.

• Forecasting: Forecasts were generated using ARIMA and ETS models.

  

  Figure 6: Forecasts generated using ARIMA (0,1,0).

  

  Figure 7: Forecasts generated using ETS (A,N,N).

Findings

The models successfully captured the trend and seasonal components of the oil sales data, providing accurate short-term forecasts.

Files

• Project 2 - RMD.Rmd: Contains the R code for the time series analysis, including data exploration, modeling, and forecasting.
• Project-2---RMD.html: An HTML report generated from the R markdown file, providing a detailed walkthrough of the time series analysis.
• oil.csv: The dataset used for time series analysis.

How to Use

1. Clone the repository: Download the project files to your local machine.
2. Open the R markdown files: Use RStudio or any compatible editor to view and run the R markdown files (.Rmd).
3. View the reports: You can directly view the HTML and PDF reports for each analysis to see the detailed results and visualizations.

Technology Stack

• R: The analysis was performed using the R programming language, leveraging its powerful libraries for data manipulation and visualization.
• R Libraries:
  • ggplot2 for data visualization.
  • dplyr for data manipulation.
  • cluster, factoextra, and dbscan for clustering algorithms.
  • forecast, tseries, and stats for time series modeling.
• Datasets: The project utilizes .csv and .xlsx datasets for analysis.

License

This project is licensed under the MIT License. See the LICENSE file for details.