Update projects/forex_algo_trading.md

projects/forex_algo_trading.md

@@ -11,26 +11,49 @@ This guide focuses on developing an algorithmic trading strategy for the EUR/USD
   - Register for an Oanda account and generate an API key.
   - Install `oandapyV20`: `pip install oandapyV20`.
 
-### 1.2 Fetching Historical EUR/USD Data
-- **Objective**: Download historical EUR/USD data suitable for ML analysis.
-- **Steps**:
-  - Use `oandapyV20` to fetch historical price data.
-  - Decide on data granularity (e.g., H1 for hourly data) based on trading strategy needs.
+## 1. Fetching Historical EUR/USD Data
 
-## 2. Data Preparation and Analysis
+### Objective
+Download historical EUR/USD data optimized for mean reversion strategy development in machine learning.
+
+### Steps
+- **API Utilization**: Employ `oandapyV20` for accessing Oanda's historical price data, focusing on capturing extensive price history to identify mean reversion opportunities.
+- **Data Granularity Decision**:
+  - For mean reversion, select granularities that balance detail with noise reduction. **H4 (4-hour)** data is a good starting point, providing insight into intraday price movements without overwhelming short-term noise.
+  - Consider also fetching **D1 (daily)** data to analyze longer-term mean reversion patterns.
+
+## 2. Data Preparation and Analysis for Mean Reversion
 
 ### 2.1 Data Cleaning and Preprocessing
-- **Objective**: Prepare the data for analysis and model training.
-- **Steps**:
-  - Handle missing values, outliers, and duplicate entries.
-  - Normalize or standardize the data if necessary.
 
-### 2.2 Exploratory Data Analysis (EDA)
-- **Objective**: Gain insights into the EUR/USD price movements and volatility.
-- **Tools**: Use `pandas` for data manipulation and `matplotlib`/`seaborn` for visualization.
-- **Steps**:
-  - Plot price movements over time.
-  - Calculate and visualize key statistics (mean, median, standard deviation).
+#### Objective
+Ensure data quality for accurate mean reversion analysis and model training.
+
+#### Steps
+- **Missing Values**: Fill or remove gaps in data to maintain consistent time series analysis.
+- **Outliers**: Identify and address price spikes that may skew mean reversion analysis.
+- **Normalization/Standardization**: Adjust data to a common scale, particularly important when combining features of different magnitudes or when data spans several years.
+
+### 2.2 Exploratory Data Analysis (EDA) for Mean Reversion
+
+#### Objective
+Identify characteristics of EUR/USD that indicate mean reversion tendencies.
+
+#### Tools and Steps
+- **Pandas for Data Handling**: Utilize `pandas` for managing time series data, crucial for chronological analysis and feature engineering.
+- **Matplotlib/Seaborn for Visualization**:
+  - **Price Movement Plots**: Visualize EUR/USD price movements with time series plots to identify cyclical patterns or periods of mean reversion.
+  - **Volatility Analysis**: Plot volatility (e.g., using ATR or standard deviation) against price to spot mean reversion during high volatility periods.
+  - **Mean Reversion Indicators**: Calculate and visualize indicators like Bollinger Bands or the Z-score (price distance from the mean), which are direct signals of potential mean reversion.
+
+#### Advanced Analysis
+- **Statistical Tests**:
+  - Conduct statistical tests like the Augmented Dickey-Fuller test to assess the stationarity of the EUR/USD series, a prerequisite for mean reversion.
+  - Use Hurst exponent analysis to differentiate between mean-reverting and trending behavior.
+
+## Next Steps: Strategy Formulation and Model Building
+- **Indicator Selection**: Beyond visual analysis, systematically select indicators that historically signal mean reversion points. Incorporate these into the feature set for ML model training.
+- **Machine Learning Models**: Experiment with models that can classify or predict mean-reverting behavior. Regression models can predict return to mean levels, while classification models can signal buy/sell opportunities based on detected mean reversion patterns.
 
 ## 3. Feature Engineering