Uber stock prediction Model Using Phased LSTM | by Sunkanmi Temidayo | Feb, 2025

Uber Applied sciences, Inc. (ticker: uber) went public on Might 10, 2019, in one of the crucial extremely anticipated IPOs of the last decade. Initially priced at $45 per share, Uber’s valuation reached roughly $82.4 billion. Nonetheless, the inventory confronted quick challenges, closing its first buying and selling day at $41.57, reflecting investor issues about profitability.

1. Pandemic and Diversification: The COVID-19 pandemic severely affected Uber’s core ride-hailing enterprise, resulting in a decline in income. Nonetheless, Uber Eats skilled exponential progress throughout this era, serving to offset losses and solidifying its position as a vital income stream.
2. Path to Profitability: Uber achieved its first adjusted EBITDA-positive quarter in Q2 2021, signaling progress towards profitability. This milestone, mixed with cost-cutting measures and operational effectivity, bolstered investor confidence in its long-term progress potential.

This challenge stands out to me as an interesting problem after I found insights about Uber’s strategic journey throughout an interview with their CEO on YouTube. This piqued my curiosity, main me to conduct a sequence of in-depth analysis efforts to higher perceive the elements driving Uber’s inventory efficiency. The mixture of exterior financial occasions, regulatory challenges, and inner diversification methods makes Uber’s story a novel case research for predictive modeling. Tackling this challenge permits me to mix my ardour for monetary evaluation and machine studying to uncover patterns in a real-world context.

The Uber Inventory Prediction Mannequin goals to investigate historic inventory worth information and predict future worth tendencies utilizing machine studying methods. The mannequin is constructed utilizing Python and well-liked information science libraries.

• Load and discover Uber inventory worth information.
• Carry out information visualization to grasp tendencies.
• Practice and consider a machine studying mannequin for inventory worth prediction.
• Generate predictions primarily based on historic patterns.

The dataset used on this challenge contains:

• Uber’s historic inventory worth information (Open, Shut, Excessive, Low, Quantity).
• Knowledge sourced from Kaggle and loaded right into a Pandas DataFrame.

• The dataset is loaded utilizing Pandas.
• The ‘Date’ column is transformed to a datetime format.
• ‘Date’ convert to African/Lagos (GMT +1) to evolve to native time when predicting.
• Lacking values are checked and dealt with.

1. Time Collection Plot: A line plot of the ‘Shut’ costs over time is generated to visualise the inventory worth pattern.
2. Seasonal Decomposition: The `seasonal_decompose` operate is used to investigate the time sequence for seasonality, pattern, and residual elements.

1. LSTM Mannequin: An LSTM community is chosen for its capability to seize long-term dependencies in time sequence information.
2. 2. Mannequin Structure: The mannequin consists of two LSTM layers adopted by two Dense layers.
   — The primary LSTM layer has 128 items and returns sequences.
   — The second LSTM layer has 64 items and doesn’t return sequences.
   — The primary Dense layer has 25 items.
   — The ultimate Dense layer has 1 unit, representing the anticipated worth. 3.
3. Knowledge Preparation: The information is cut up into coaching and testing units. The coaching set is additional divided into enter sequences (X_train) and goal values (y_train).
4. Mannequin Compilation: The mannequin is compiled utilizing the ‘adam’ optimizer and ‘mean_squared_error’ loss operate.
5. Mannequin Coaching: The mannequin is educated for 50 epochs with a batch dimension of 32.

1. Check Knowledge Preparation: The check information is ready in an analogous technique to the coaching information, creating enter sequences (X_test) and goal values (y_test).
   2. Mannequin Analysis: The mannequin is evaluated on the check information utilizing the `consider` methodology, calculating the check loss.
   3. Prediction: The mannequin is used to foretell inventory costs for the check information.
   4. Visualization: The precise and predicted costs are plotted collectively for comparability.
   5. Comparability Desk: A desk is created to show the precise and predicted costs side-by-side.

• The mannequin offers a fundamental forecast of Uber’s inventory costs.
• Visualization helps establish tendencies and patterns in inventory worth motion.

The LSTM mannequin demonstrates the potential for predicting inventory costs. The outcomes present that the mannequin is ready to seize the overall pattern of the inventory worth motion. Nonetheless, additional enhancements will be explored by tuning hyperparameters, utilizing completely different architectures, or incorporating extra options.

Undertaking: Go to GitHub