KSPI Stock Forecast

Outlook: KSPI is assigned short-term Ba3 & long-term B1 estimated rating.
AUC Score : What is AUC Score?
Short-term Tactic¹ :
Dominant Strategy :
Time series to forecast n: for Weeks²
ML Model Testing : Multi-Task Learning (ML)
Hypothesis Testing : Sign Test
Surveillance : Major exchange and OTC

¹Short-term revised.

²Time series is updated based on short-term trends.

Key Points

Kaspi.kz ADRs are predicted to experience significant growth driven by its dominant position in the rapidly expanding Kazakhstani digital ecosystem, encompassing payments, e-commerce, and fintech. However, potential risks include increasing competition from local and international players, regulatory changes impacting its business model, and geopolitical instability in its operating region which could affect investor sentiment and operational continuity.

About KSPI

This exclusive content is only available to premium users.

KSPI American Depository Shares Stock Forecast Model

Our proposed machine learning model for forecasting Joint Stock Company Kaspi.kz American Depository Shares (KSPI) leverages a hybrid approach combining time-series analysis with sentiment analysis of news and social media data. The core of our time-series component will utilize Long Short-Term Memory (LSTM) networks, a type of recurrent neural network adept at capturing complex temporal dependencies and patterns within historical stock price movements. We will incorporate a comprehensive set of technical indicators derived from historical trading data, such as moving averages, relative strength index (RSI), and MACD, as input features for the LSTM. This allows the model to learn from established charting patterns and momentum indicators. Concurrently, a natural language processing (NLP) module will be employed to extract sentiment scores from a diverse corpus of financial news articles, analyst reports, and relevant social media discussions pertaining to Kaspi.kz and the broader fintech and e-commerce sectors. This sentiment data will be integrated as an additional feature set, enabling the model to account for the impact of market sentiment and news events on stock performance.

The development process involves rigorous data preprocessing, including handling missing values, feature scaling, and segmenting the data into training, validation, and testing sets. We will explore various architectural configurations for the LSTM network, including the number of layers, units per layer, and dropout rates, to optimize predictive accuracy. Feature engineering will play a crucial role, focusing on creating meaningful combinations of technical and sentiment-derived features. For instance, interactions between specific sentiment scores and periods of high trading volume might reveal important predictive signals. The NLP component will undergo fine-tuning on domain-specific financial language to ensure accurate sentiment classification and intensity measurement. We will employ a robust evaluation framework using metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared to assess the model's performance on unseen data, ensuring it generalizes well beyond the training set and avoids overfitting.

The anticipated output of this model is a probabilistic forecast of future KSPI stock price movements over short to medium-term horizons. This will provide valuable insights for investment decisions, risk management, and portfolio optimization for stakeholders. The model's architecture is designed for adaptability, allowing for continuous retraining with updated data to maintain its predictive power as market dynamics evolve. Furthermore, we will develop a system for interpreting model outputs, providing explanations for significant forecast deviations through feature importance analysis. This transparency is crucial for building trust and enabling informed strategic planning for Joint Stock Company Kaspi.kz and its investors. The objective is to deliver a reliable and actionable forecasting tool that contributes to superior investment outcomes.

ML Model Testing

F(Sign Test)^6,7= $\begin{array}{cccc}{p}_{\mathrm{a}1}& p_{\mathrm{a}2}& \dots & p_{1n}\\ & ⋮\\ p_{j1}& p_{j2}& \dots & p_{jn}\\ & ⋮\\ p_{k1}& p_{k2}& \dots & p_{kn}\\ & ⋮\\ p_{n1}& p_{n2}& \dots & p_{nn}\end{array}$ X R(Multi-Task Learning (ML))^3,4,5 X S(n):→ 16 Weeks $\overrightarrow{R}=\left(r_1,r_2,r_3\right)$

n:Time series to forecast

p:Price signals of KSPI stock

j:Nash equilibria (Neural Network)

k:Dominated move of KSPI stock holders

a:Best response for KSPI target price

 

For further technical information as per how our model work we invite you to visit the article below: 

How do KappaSignal algorithms actually work?

KSPI Stock Forecast (Buy or Sell) Strategic Interaction Table

Strategic Interaction Table Legend:

X axis: *Likelihood% (The higher the percentage value, the more likely the event will occur.)

Y axis: *Potential Impact% (The higher the percentage value, the more likely the price will deviate.)

Z axis (Grey to Black): *Technical Analysis%

This exclusive content is only available to premium users.

Rating	Short-Term	Long-Term Senior
Outlook	Ba3	B1
Income Statement	Caa2	B3
Balance Sheet	Ba3	B1
Leverage Ratios	Baa2	B1
Cash Flow	Baa2	B3
Rates of Return and Profitability	Caa2	B1

*Financial analysis is the process of evaluating a company's financial performance and position by neural network. It involves reviewing the company's financial statements, including the balance sheet, income statement, and cash flow statement, as well as other financial reports and documents.
How does neural network examine financial reports and understand financial state of the company?

References

1. Scott SL. 2010. A modern Bayesian look at the multi-armed bandit. Appl. Stoch. Models Bus. Ind. 26:639–58
2. Athey S, Imbens G. 2016. Recursive partitioning for heterogeneous causal effects. PNAS 113:7353–60
3. Athey S, Blei D, Donnelly R, Ruiz F. 2017b. Counterfactual inference for consumer choice across many prod- uct categories. AEA Pap. Proc. 108:64–67
4. LeCun Y, Bengio Y, Hinton G. 2015. Deep learning. Nature 521:436–44
5. Zou H, Hastie T. 2005. Regularization and variable selection via the elastic net. J. R. Stat. Soc. B 67:301–20
6. S. J. Russell and A. Zimdars. Q-decomposition for reinforcement learning agents. In Machine Learning, Proceedings of the Twentieth International Conference (ICML 2003), August 21-24, 2003, Washington, DC, USA, pages 656–663, 2003.
7. A. Tamar, Y. Glassner, and S. Mannor. Policy gradients beyond expectations: Conditional value-at-risk. In AAAI, 2015