PredicTrade

Purpose of Application

This is an educational project designed to teach me about machine learning and the various financial and non-financial data points that affect stock prices. I want to better understand the work that goes into quantitative firms' algorithms for predicting stock prices and automating trading decisions.

The goal is to create a working project with solid performance metrics, backtested against both my personal portfolio and randomly selected stocks from major indices (Dow Jones, Nasdaq, S&P 500, Russell 2000, and potentially other US and international markets). Three levels - My Portfolio (Buy More, Sell, Hold) Assortment of Tracked Stocks (based off of features selected, predict price and get accuracy metrics, then give buy/no buy) Assortment of Random Stocks (based off of features selected, predict price to get accuracy metrics, then give buy/no buy)

Additionally, I plan to investigate whether the stock market has become disconnected from the real economy by analyzing market performance against common household economic indicators.

Planned Feature Implementation

✅ Completed Features

1. Sentiment Analysis - Orthogonal Fear-Greed Index separating Sentiment from Financial Data
   • Google Trends (35%): 70/30 US/world weighted search sentiment
   • News Sentiment (30%): Pygooglenews + TextBlob analysis
   • Professional Sentiment (20%): Finnhub financial news + social sentiment
   • Market Context (15%): VIX-based market fear from FRED API

🎯 Next Priority: Economic Context Engine

Individual FRED Economic Indicators for Testing & Research:

Real Economy Health Indicators

• GDP Growth Rate (GDPC1) - Quarterly real GDP growth
• Unemployment Rate (UNRATE) - Monthly unemployment percentage
• Consumer Price Index (CPIAUCSL) - Monthly inflation measure
• Personal Income Growth (PI) - Monthly personal income changes
• Housing Price Index (CSUSHPISA) - Monthly housing price trends

Household Economic Indicators

• Consumer Confidence Index (UMCSENT) - Monthly consumer sentiment
• Personal Savings Rate (PSAVERT) - Monthly household savings percentage
• Consumer Credit (TOTALSL) - Total consumer loans outstanding
• Retail Sales Growth (RSAFS) - Monthly retail sales changes

Financial Conditions Indicators

• Federal Funds Rate (FEDFUNDS) - Monthly Fed policy rate
• 10-Year Treasury Yield (GS10) - Daily 10-year bond yield
• 3-Month Treasury Yield (GS3M) - Daily 3-month bond yield
• Term Spread (GS10 - GS3M) - Yield curve indicator
• Dollar Index (DTWEXBGS) - Trade-weighted USD strength

Market Context Indicators (Existing)

• VIX (VIXCLS) - Market volatility/fear index ✅ (Already implemented)

📊 Phase 3: Technical Analysis Engine

• Price-based indicators (SMA, EMA, RSI, MACD, Bollinger Bands)
• Volume-based indicators (OBV, Volume Rate of Change)
• Momentum indicators (Stochastic, Williams %R)

💰 Phase 4: Fundamental Analysis Engine

• Financial ratios (P/E, P/B, P/S, EV/EBITDA, ROE, ROA)
• Growth metrics (Revenue, Earnings, FCF growth)
• Financial health metrics (Debt ratios, liquidity ratios)

Data Sources

1. yfinance: base api, scrapes for its data so will be used for fallback method primarily if I hit an api endpoint and I get an error indicating that I have used too many calls.
   1. https://ranaroussi.github.io/yfinance/reference/yfinance.stock.html
   2. https://ranaroussi.github.io/yfinance/reference/yfinance.financials.html
   3. https://ranaroussi.github.io/yfinance/reference/yfinance.analysis.html
2. Finnhub: One of three primary sources for data, mostly US focused.
   1. https://finnhub.io/docs/api/company-basic-financials contains 10DayAverageTradingVolume", "52WeekHigh", "52WeekLow", "52WeekLowDate", "52WeekPriceReturnDaily", "beta"
   2. https://finnhub.io/docs/api/earnings-calendar\ (1 month of historical earnings and new updates for free tier) contains "date", "epsActual", "epsEstimate", "quarter", "revenueActual", "revenueEstimate", "year"
   3. https://finnhub.io/docs/api/quote contains "c" (current price), "h" (high price of day), "l" (low price of day), "o" (open price of day), "pc" (previous close), "t" (timestamp)
   4. https://finnhub.io/docs/api/company-peers returns list of peers, (grouping optional) but can be grouped as followed: sector, industry, subIndustry. Default to subIndustry
   5. https://finnhub.io/docs/api/recommendation-trends contains "buy", "hold", "period", "sell", "strongBuy", "strongSell", "symbol"
3. Tiingo: One of three primary sources of data, generous limits for requests (request in CSV format for ability to train models and not need to continuously API and waste calls)
   1. https://www.tiingo.com/documentation/fundamentals (3 Years of the DOW 30 tickers are available for free)
      1. API Endpoint: https://api.tiingo.com/tiingo/fundamentals/(tickerHere)/daily contains Date, Market Cap, Enterprise Value, P/E Ratio, P/B Ratio, Trailing PEG Ratio (1Y)
      2. API Endpoint: https://api.tiingo.com/tiingo/fundamentals/(tickerHere)/statements contains Date, Quarter, Year, Statement Data
   2. https://www.tiingo.com/documentation/end-of-day
      1. API Endpoint: https://api.tiingo.com/tiingo/daily/(tickerHere)/prices?startDate=2012-1-1&endDate=2016-1-1&format=csv&resampleFreq=monthly contains Date, Open, High, Low, Close, Volume, Adj, Open, Adj, High, Adj. Low, Adj. Close, Adj. Volume, Dividend, Split
   3. https://www.tiingo.com/documentation/iex (could be a good cross-check for alpaca?)
   4. https://www.tiingo.com/documentation/news (maybe should look into adding for sentiment analysis?)
4. Alpaca Markets: Secondary source of data, generous free tiers for requests (including live data 500 requests per day), only downside free tier is limited to IEX exchange in US
   1. https://docs.alpaca.markets/reference/stockbarsingle-1
   2. https://docs.alpaca.markets/reference/stockbars
   3. https://docs.alpaca.markets/reference/corporateactions-1
   4. https://docs.alpaca.markets/reference/stocksnapshots-1 (might be the one I need combined with 3)
5. FRED API: Financial St. Louis FED Market API, One of three primary sources of data that provides official federally sourced free data to potentially calculate a Fear-Greed Index for specific stocks
6. (Not an API, but congressional trades) -> https://pelositracker.app/portfolios, would be a good lagging indicator of stock potential?