Autoregressive Model Analysis and Multistep Forecast of Apple Stock Data

Table of Contents

• Overview
• System Prerequisites
• Getting Started
• Project Structure

---

Overview

In this project, Apple stock data was analyzed using time series econometrics methods. The final output of this project is the LaTeX file paper.pdf that roughly describes the analysis steps as well as the figures/tables and deals with the question to what extent it is possible to interpret the results.

In this project, I:

• fit several AR processes to Apple's historical stock data.
• compared model performance to identify the best fitting AR process.
• evaluated the ability of the best model to perform multistep forecasts.
• investigated the extent to which it is possible to use the AR process for analysis.
• provided analysis and plots to visualize both the model fit and forecasting performance.

---

System Prerequisites

To make sure that the project works on your device it is necessary to have installed Python, a modern LaTeX distribution, Git, and if applicable a text editor. For a more detailed explanation see the documentation.

---

Getting Started

First one needs to clone the repository:

git clone https://github.com/iame-uni-bonn/final-project-Lenr4.git

Next, navigate to the project root and create and activate the environment:

mamba env create lennart_epp
conda activate lennart_epp

After the environment is activated, one can run the project by:

pytask

🛑 Caution: If there were any trouble with kaleido on windows you need to use this workaround:

pip install kaleido==0.1.0.post1

---

Project Structure

The Project is structured into three different parts.

• bld: The Build directory cointaing all output files.

  • plots: Top 3 AR models for fitting (1 step forecast), ACF, Multistep forecast; all as interactive .html and .pdf files.
  • forecasts: Multistep forecast using AR(1) as .pkl file.
  • data: Cleaned Apple data as .pkl file.
  • memory: .pkl file of ACF and .tex files of Hurst and ADF statistics.
  • models: .pkl file of all AR models and .tex file with top model statistics.

• src: The Source directory containing all python files needed for the analysis.

  • data: CSV file containing the raw data for reproducibilty.
  • data_management: Python files for cleaning and downloading the data from Yahoo Finance.
  • analysis: Python files which analyse the data.
  • final: Python files which plot the results.

• tests: The Test directory containing all python files which are used for testing.

  • data_management: Python files for testing the data management steps.
  • analysis: Python files for testing the analysis steps.

---