FoudaLens uses artificial intelligence to generate daily price predictions for Egyptian stocks. The AI system combines multiple data sources — technical analysis, fundamental data, market sentiment, and historical patterns — to produce a prediction that includes a direction (up/down/neutral), a confidence level, and an expected magnitude. This article explains how the system works and how to interpret its output.
The prediction pipeline has three stages: (1) Feature engineering — the system computes hundreds of features from raw market data, including technical indicators (RSI, MACD, Bollinger, ADX, etc.), fundamental ratios (PE, PB, ROE), volume patterns, and inter-stock correlations. (2) Model inference — these features are fed into a machine learning model that has been trained on years of Egyptian market data to predict short-term price movements. (3) Post-processing — the raw prediction is calibrated, combined with sentiment data, and converted into a human-readable format with confidence levels.
Confidence levels indicate how certain the model is about its prediction. High confidence (above 75%) means the model's inputs are strongly aligned — technical, fundamental, and sentiment signals all point in the same direction. Medium confidence (50-75%) means most signals agree but some are conflicting. Low confidence (below 50%) means signals are mixed or the model encounters a situation outside its training distribution. Higher confidence predictions are historically more accurate but not guaranteed.
Technical vs. sentiment analysis: The AI weighs both quantitative technical signals and qualitative sentiment data. Technical analysis is derived from price and volume data using the same indicators described in other articles. Sentiment analysis incorporates news data, social media mentions, and broader market conditions. During normal market conditions, technical analysis dominates the prediction. During major news events (policy changes, earnings surprises, currency moves), sentiment analysis receives higher weight.
The Brain system continuously evaluates prediction accuracy and adjusts model weights. When predictions for certain types of stocks or market conditions are consistently wrong, the Brain identifies the failure pattern and suggests weight adjustments. These adjustments go through an approval process before being applied. This self-learning loop means the model improves over time as it encounters new market scenarios.
Prediction accuracy metrics: FoudaLens tracks multiple accuracy measures: (1) Directional accuracy — what percentage of up/down predictions were correct. (2) Magnitude accuracy — how close the predicted price change was to the actual change. (3) Confidence calibration — are predictions labeled 80% confidence actually correct 80% of the time? No prediction model is perfect — typical directional accuracy for stock predictions ranges from 55-65%. The value is in the statistical edge over random guessing, compounded over many trades.
How to use AI predictions: (1) Use them as ONE input among many — never trade based solely on a prediction. (2) Weight predictions by confidence — a high-confidence prediction deserves more attention than a low-confidence one. (3) Combine with your own technical and fundamental analysis. (4) The prediction horizon is short-term (1-5 trading days) — do not use them for long-term investment decisions. (5) Track the prediction history for stocks you follow to calibrate your trust. (6) During high market volatility, all predictions become less reliable — reduce reliance during market chaos. This is not financial advice.
This content is for educational purposes only and does not constitute financial advice. Always do your own research before making investment decisions.