ADS Capstone Chronicles Revised

5

effectiveness with real-time data is not demonstrated. Another concern is the high transaction costs associated with the trades. After 450 rounds of training, the transaction fees exceeded $5,000. While this may be manageable for major broker companies, it raises questions about its feasibility for everyday traders. Testing the model on cryptocurrencies with lower transaction costs, such as Solana or Cardano, might be worthwhile to see if it remains effective and affordable for individual traders. 4 Data Architecture Figure 1 shows the project ’ s overall architecture, outlining the end-to-end process of extracting, processing, and analyzing cryptocurrency data. A Python script extracts various values from the API, including time, high, low, open, close, volume, and more. These values pertain to 10 cryptocurrencies, including Ethereum, Bitcoin, Litecoin, and others. Once extracted, the data is saved as CSV files — one for historical price values and another for more recent timeframes.

After extraction and storage in the raw data folder, exploratory data analysis (EDA) is performed in the project analysis notebook. Upon completion of EDA, the cleaned data is stored in the respective cleaned folder. This cleaned data is then used for further preparation for modeling, such as time series modeling. Upon completing the modeling, forecasted future values are generated based on the training data and are compared with the values in the test set to evaluate performance on unseen data, resulting in the creation of the forecasted versus actual file. Finally, the results are displayed through a Streamlit application, which presents the most compelling trading trend analyses. The application combines data processing scripts and visualization logic to create an interactive user interface, enabling users to explore the forecast results and compare them against unseen data.

Figure 1 Data Life Cycle Graph

extracted from this endpoint included ‘ time, ’ ‘low,’ ‘high,’ ‘open , ’ ‘close,’ and ‘volume . ’ From these features, new columns such as ‘price_change , ’ ‘average_price , ’ ‘volatility , ’ and ‘product_id’ were engineered during the

4.0.1 Data Extraction

For this analysis, data were extracted from the Coinbase API endpoint candles. The features

72