Evaluation of Zero-Shot Transformer Models for Real-Time Churn Intent Classification

Main Objective: 

To implement a non-linear NLP (Natural Language Processing) pipeline capable of categorizing unstructured customer feedback without prior task-specific training.

Model Output :

Text: "I think the subscription has become too expensive for what it is." Prediction: Churn Risk (Confidence: 0.58).

Text: "I love my plan, the network is great everywhere!" Prediction: Support Request (Confidence: 0.36).

Text: "My contract is expiring soon and I'm looking at the competition." Prediction: Support Request (Confidence: 0.52).

Key Takeaways :

For the first result (0.58): > "The model identified a pricing pain point. In natural language processing, 'expensive' is a strong predictor for customer attrition, hence the 'Churn Risk' label.

For the second result (0.36): > "This is a false positive due to a low confidence score. Because the model didn't have a 'Positive' or 'Satisfied' category to choose from, it defaulted to 'Support Request' with very low certainty. We call this a forced choice.

For the third result (0.52): > "The model correctly sensed an intent to switch. Mentioning 'competition' and 'contract expiration' triggers a high probability for a retention-related support ticket."

🎯 The detailed methodology and results can be accessed through this link:

👉click here now! :  https://www.geeksforgeeks.org/

The Abdi-Basid Courses Institute (tABCi)

@ (c)2025 Abdi-Basid ADAN [abdi-basid@outlook.com]

Industrial Reliability Study: Predicting APS System Failures in Scania Trucks

Description 

The APS (Air Processing System) is critical for the operation of Scania trucks. APS failures can lead to significant industrial costs and production downtime. Historical sensor data enables the analysis of system behavior and early detection of anomalies before they escalate into critical failures.

In this study, a Random Forest model was trained to detect APS system failures, with special attention given to the highly imbalanced nature of the dataset by applying class weighting to the rare positive instances. Model performance was evaluated using standard metrics including precision, recall, F1-score, and confusion matrices, while the associated industrial cost was calculated based on the impact of false positives and false negatives. To further optimize failure detection and reduce total costs, XGBoost was optionally employed. Additionally, feature importance analysis was conducted to identify the most critical sensors influencing APS failure predictions, with the top features visualized through horizontal bar charts to provide interpretable insights for industrial decision-making.

Figure 1 illustrates the confusion matrix of the Random Forest model applied to APS failure detection under a highly imbalanced class distribution. The model correctly classifies the majority of non-failure cases, as reflected by the large number of true negatives and the very low number of false positives. This indicates a strong capability to avoid unnecessary maintenance actions. However, a non-negligible number of APS failures are misclassified as normal operations (false negatives), which represent critical cases from an industrial perspective due to their high associated cost. This result emphasizes the importance of cost-sensitive learning and motivates the use of alternative models, such as XGBoost, to further reduce false negatives and optimize industrial cost. In the confusion matrix, the value 15,607 corresponds to true negatives, that is, trucks that do not have an APS system failure and are correctly identified as such by the model, demonstrating its ability to avoid unnecessary maintenance interventions.

Figure 2 presents the top 15 most influential sensor variables used by the Random Forest model to predict APS failures. The results indicate that the prediction is driven by a limited subset of sensors, with aa_000 being the most dominant feature, followed by ci_000, ck_000, and dn_000. This suggests that APS failures are strongly associated with specific operational measurements rather  than uniformly across all sensors. The concentration of importance among these variables highlights their potential relevance for targeted monitoring and preventive maintenance strategies, as focusing on key sensors could improve fault detection efficiency while reducing system complexity.

Conclusion

The experimental results confirm that the proposed machine learning approach is effective for detecting failures of the Air Pressure System (APS) in heavy-duty trucks. The Random Forest model achieved a high overall accuracy (≈99%) and a strong precision for the negative class, indicating reliable identification of non-failure cases. However, the recall for APS failures remained moderate (≈58–61%), highlighting the intrinsic difficulty of detecting rare failure events in highly imbalanced industrial datasets.

The integration of class weighting significantly reduced the number of false negatives, which are associated with the highest industrial cost. Using the defined cost function, the Random Forest model resulted in a total industrial cost of approximately 74,000–78,000 units, demonstrating a meaningful improvement over non–cost-aware baselines. Furthermore, the XGBoost model substantially outperformed Random Forest in cost optimization, reducing the total industrial cost to approximately 29,850 units, primarily by further decreasing missed APS failures.

Feature importance analysis revealed that a limited set of sensor variables (e.g., aa_000, ci_000, ck_000, dn_000) consistently contributed most to the predictive performance. This suggests that APS degradation can be detected through specific operational patterns captured by onboard sensors. Overall, these results validate the relevance of cost-sensitive learning for industrial reliability studies and demonstrate the practical value of data-driven predictive maintenance in intelligent transportation systems.

🎯 The detailed methodology and results can be accessed through this link:

👉click here now! :  https://github.com/

  

The Abdi-Basid Courses Institute (tABCi)

@ (c)2025 Abdi-Basid ADAN [abdi-basid@outlook.com]

NCAA Match Prediction Script Using Logistic Regression

This script aims to predict the probability that a college basketball team wins against another in a match. It leverages historical data on teams, regular season results, and tournament seeds to build a predictive model.

Purpose :

- Understand how to transform raw match data into features usable

by a machine learning model.

- Build a supervised model capable of predicting the match winner.

- Evaluate the model using standard metrics (log loss, ROC-AUC)

and apply it to tournament simulations.

Data :

- "teams": team information (TeamID, TeamName, first and last

Division 1 season)

- "results": regular season match results (winning team, losing team,

score, match day)

- "seed_round_slots": information on tournament seeds and match slots

Variables:

- "team_stats": number of wins and losses per team per season

- "match_data": prepared match dataset for model training

- "X", "y": features and target for training

- "model": trained logistic regression model

- "matchup_example": sample tournament matches for prediction

Model:

- Logistic Regression

- It is supervised because it learns from labeled data: each historical

match has a label "1" if Team1 wins, "0" otherwise.

- Suitable for binary classification and allows estimating the probability

of a team winning.

Objectives:

1. Load the necessary CSV files.

2. Compute wins and losses for each team and season.

3. Create a match dataset ready for training.

4. Normalize the data and split into training and test sets.

5. Train a supervised Logistic Regression model.

6. Evaluate the model using log loss and ROC-AUC.

7. Prepare a sample tournament matchup and predict win probabilities.

🎯 The detailed methodology and results can be accessed through this link:

👉click here now! :  https://github.com/

    

The Abdi-Basid Courses Institute (tABCi)

@2025 Abdi-Basid ADAN

Innovative Discoveries on the Laws of Cosmology and Humanity: New Perspectives on the World

A B S T R A C T : This study proposes an innovative framework for exploring the laws of human cosmology by modeling consciousness as a dynamic entity integrating six fundamental sensory dimensions (taste, vision, smell, hearing, internal/external touch, and soul). Inspired by the treatise Mystères Dévoilés : Vérités Novatrices – Révélations des Lois de la Cosmologie de l’Homme (Adan, 2025), the aim is to link consciousness to parallel realities through sensory vibrations, while emphasizing the supreme position of human beings within nature. An observational and contemplative methodology was employed, combining a lit- erature review in neuroscience and philosophy with immersive observations in African and European societies (2020–2025), along with meditative reflections accounting for variations related to age, gender, and cultural context. The findings indicate that consciousness emerges as a dual transverse wave, evolving from a fetal pseudo-consciousness to a mature postnatal form, with passive anomalies such as hypnosis or hallucinations. It interacts with the environment through six senses, generating vibrations that give rise to parallel realities; neurons function as a bridge between body and soul, whereas sleep illustrates vital discontinuity. At the cosmic scale, consciousness links life and the universe, influenced by time and predestination. These discoveries offer a unified model between the material and immaterial realms, with implications for scientific philosophy, addiction prevention, and the preservation of human dignity as a supreme ethical responsibility.

Keywords: Adanian Theories Sensory Vibrations Parallel Realities Soul and Gravity Consciousness Human Cosmology Scientific Philosophy

🎯 The detailed methodology and results can be accessed through this link:

👉click here now! : 

https://archive.org

Adan, A.-B. (2025). Innovative Discoveries on the Laws of Cosmology and Humanity: New Perspectives on the World. https://doi.org/10.5281/zenodo.18035321

MystèresDévoilés :Vérités NovatricesRévélations des Lois de la Cosmologiede l’Homme : Nouvelles Visions du Monde

Cet ouvrage, à la dimension d’un traité philosophique, au sens propre, est amené avant tout, de dérouler une variété des structures en concepts à travers des véracités fondés sur un ensemble des raisonnements déductifs, observationnels et logiques, entretenant des perspectives transversales, longitudinales et pyramidales pour mettre en œuvre non seulement des solutions transparentes à l’égard des problèmes complexes, mais aussi formuler des sous-ensembles des théorèmes, propriétés, axiomes, postulats, assomptions et hypothèses, qui s’agglomèrent pour édifier à leur tour les piliers des théories Adaniennes.

En termes simples, cet ouvrage initie l’universalité de la scientificité en philosophie, en mettant sur pied des préceptes dont la plupart sont novatrices, afin d’émanciper la multiplicité en recherche dans le domaine de la philosophie et les disciplines adjacentes et d’autres part, redéfinir les bases connues en principes de la littérature en philosophie. Les théories développés et améliorés dans cet ouvrage sont à la fois amplificatrice et simplificatrice pour une accessibilité à un large public leur offrant une vue d’ensemble à 360°.

🎯 The detailed methodology and results can be accessed through this link:

👉click here now! : 

https://www.amazon.com Or  https://www.amazon.com

@ (c)2025 Abdi-Basid ADAN [abdi-basid@outlook.com]

Modeling Rainfall Variability and Extremes in Djibouti under CMIP5, CMIP6, and CORDEX Scenarios

Description:

This study investigates historical and projected rainfall patterns at Djibouti Airport using observational datasets (CHIRPS, ERA5Land) and climate models, including CanESM2, CanESM5, and regional CORDEX downscaled simulations. The research evaluates model performance through statistical metrics, such as Taylor diagrams and extreme precipitation indices (ETCDDI), and examines projected changes under multiple climate scenarios (RCP4.5, RCP8.5, SSPs). The results provide insights into future rainfall variability, trends in extreme events, and potential drought and wet periods, supporting climate risk assessment and adaptation planning in the Horn of Africa.

Figure 1. Historical CHIRPSv.2 and ERA5Land annual rainfall at Djibouti airport station compared to corrected rainfall from CanESM2 and RCM (AFR22-CanRCM4, AFR44-SMHI-RCA4, AFR44-CanRCM4, AFR44-UQAM-CRCM5) over the period 1980-2005 and Taylor Diagram for performance comparison (left panels).The radial coordinate indicates the variance ratio between the observation and the satellite data.

Figure 2. Projected rainfall changes relative to the baseline period 1953–2021 based on CanESM2-CORDEX, the downscaled CanESM2-CMIP5 and CanESM5-CMIP6 using Observation, CHIRPS and ERA5Land datasets. Colored shaded areas represent the areas of uncertainty (standard deviation) for the scenarios RCP4.5, RCP.8.5 for CMIP5 and SSP2-RCP4.5and SSP5-RCP8.5 for CMIP6 (right panels). The solid sphere in the boxplot represents the mean, and the interquartile range spans from Q1 to Q3 within the box square. Horizontal lines above and below denote the minimum and maximum. Irregular dotted indicate extreme values (left panels).

Table 1. Overall change of rainfall for future climate generated using CanESM2, AFR44 and CanESM5 models in RCP 4.5 and RCP 8.5 scenarios.

Figure 3. Historical and projected average monthly rainfall at Djibouti airport station (1953-2021) using statistical downscaling of CanESM2 and CanESM5 and corrected regional model CORDEX from 2006–2100.

Table 2. Trend of ETCDDI indices for extreme precipitation (historical (1980-2017) and projected with AFR44.CanRCM4.CHIRPS.RCP45 (2018-2099) at the Djibouti airport station and detection of the stationarity period with the Pettitt test (Pettitt, 1979) and standard normal homogeneity (SNHT, Alexandersson 1986).

Figure 4. Projected Interannual variations (line) and linear trends (dash) in mean annual 3-month RAI at Djibouti airport station from downscaled CMIP5, CMIP6 and CORDEX based on observation, CHIRPS and ERA5Land datasets (right panels) and Characteristics of drought (left panels).

Figure 5. Variation of ETCDDI indices for projected extreme precipitation (2018-2099) at the Djibouti airport station with the CanESM2 (CHIRPS.RCP45 and ERA5Land. RCP85), CanESM5 (CHIRPS. RCP45 and ERA5Land. RCP45) and CORDEX (CHIRPS. RCP45 and ERA5Land. RCP45) models.

Figure 5. Projected rainfall changes relative to the baseline period 1953–2021 based on the downscaled CanESM2-CMIP5 and CanESM5-CMIP6 using Observation, CHIRPS and ERA5Land datasets. Colored shaded areas represent the areas of uncertainty (standard deviation) for the scenarios RCP1.9, RCP.2.6 and RCP.7.0 for CMIP5 and SSP1-RCP2.6, SSP3-RCP7.0 and SSP5-RCP8.5 for CMIP6. The solid sphere in the boxplot represents the mean, and the interquartile range spans from Q1 to Q3 within the box square. Horizontal lines above and below denote the minimum and maximum. Irregular dotted indicate extreme values.

Table 3. Evaluation of precipitation projections (very short term) from 2006 to 2021 under the RCP 4.5 scenario simulated at the Djibouti airport station by Canadian Earth System models derived from CMIP5, CORE-CORDEX and CMIP6.

Table 4. Evaluation of precipitation projections (very short term) from 2006 to 2021 under the RCP 8.5 scenario simulated at the Djibouti airport station by Canadian Earth System models derived from CMIP5, CORE-CORDEX and CMIP6.

Table 5. Overall change of average rainfall for future climate generated using CanESM2, AFR44 and CanESM5 model RCP 4.5 scenarios.

RAINFALL TRESHOLD    (Djibouti Case)

·         Q12.5%    less than     35.75 mm/annum    (Dry events)

·         Q87.5%    more than   291.60/annum         (Wet events)

·         Q50%       less than      1 mm/month          (Dry events)

·         Q90%       more than   32.72 mm/month    (Wet events)

·         Light rainfall (0–0.2(q95) mm/day)

·         Moderate rainfall events (0.2–8(q99) mm/day)

·         Heavy rainfall events (> 8 mm/day).

Abdi-Basid ADAN, 2024

The Abdi-Basid Courses Institute (tABCi)