The American Journal of Interdisciplinary Innovations and Research
66
https://www.theamericanjournals.com/index.php/tajiir
Type
Original Research
PAGE NO.
66-73
10.37547/tajiir/Volume07Issue08-07
OPEN ACCESS
SUBMITED
25 July 2025
ACCEPTED
04 August 2025
PUBLISHED
18 August 2025
VOLUME
Vol.07 Issue 08 2025
CITATION
Priyank Tailor. (2025). Automating Financial Risk Assessment Using NLP
and Machine Learning. The American Journal of Interdisciplinary
Innovations and Research, 7(8), 66
–
73.
https://doi.org/10.37547/tajiir/Volume07Issue08-07
COPYRIGHT
© 2025 Original content from this work may be used under the terms
of the creative commons attributes 4.0 License.
Investi
Automating Financial Risk
Assessment Using NLP and
Machine Learning
Priyank Tailor
Data Scientist / AI Researcher Jersey City, NJ, USA
Abstract
- Financial risk assessment is a critical function
within the financial industry, encompassing the
identification, measure- ment, and mitigation of various
risks such as credit risk, market risk, operational risk, and
liquidity risk. Traditional methods often rely on
quantitative models built upon struc- tured numerical
data, which, while effective, frequently over- look the
vast amount of unstructured information available in
financial documents. This paper explores the integration
of Natural Language Processing (NLP) and Machine
Learn- ing (ML) techniques to automate and enhance
financial risk assessment. We propose a comprehensive
framework that leverages NLP to extract meaningful
insights from diverse unstructured textual data sources,
including financial news, company reports, social media,
and regulatory filings. These extracted features,
combined with traditional quantitative data, are then
fed into advanced machine learning models to provide
more accurate, timely, and holistic risk evalua- tions.
Our approach aims to overcome the limitations of
existing models by providing a more accurate, timely, and
in- terpretable solution for financial market analysis,
ultimately leading to more robust decision-making and
improved fi- nancial stability. We demonstrate how NLP
can identify early warning signals, detect emerging risks,
and provide a nuanced understanding of market
sentiment and corporate health, which are often missed
by purely numerical analyses. The integration of ML
models further allows for the identi- fication of complex
patterns and predictive capabilities that enhance the
overall risk assessment process.
Keywords:
Financial Risk Assessment, Natural Language
Processing, Machine Learning, Credit Risk, Market Risk,
Operational Risk, Liquidity Risk, Unstructured Data,
Financial News, Company Reports, Regulatory Filings,
The American Journal of Interdisciplinary Innovations and Research
67
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
Sentiment Analysis, Predictive Modeling.
1
. Introduction
The financial industry operates in an environment
charac- terized by inherent uncertainties and a
multitude of risks. Effective financial risk assessment is
paramount for main- taining stability, ensuring
regulatory
compliance,
and
making
informed
investment and lending decisions. Traditionally, risk
assessment has been heavily reliant on quantitative
mod- els that analyze structured data such as balance
sheets, in- come statements, and historical market
prices. While these models provide valuable insights into
quantifiable risks, they
often fall short in capturing the qualitative, nuanced,
and often forward-looking information embedded
within unstruc- tured textual data. This oversight can
lead to incomplete risk profiles and delayed responses
to emerging threats.
The proliferation of digital information has led to an
explo- sion of unstructured textual data relevant to
financial markets. This includes, but is not limited to,
financial news articles, analyst reports, company annual
reports (10-K, 10-Q), earn- ings call transcripts, social
media discussions, and regulatory announcements.
These diverse sources contain critical qual- itative
information about macroeconomic trends, geopoliti- cal
events, corporate governance issues, technological dis-
ruptions, and market sentiment
–
all of which
significantly influence financial risk. The challenge lies in
efficiently pro- cessing and extracting actionable
intelligence from this vast and complex data landscape.
Natural Language Processing (NLP) has emerged as a
pow- erful set of techniques capable of understanding,
interpreting, and generating human language. Its
application in finance has gained significant traction,
moving beyond simple keyword searches to
sophisticated sentiment analysis, entity recogni- tion,
topic modeling, and text summarization. By applying
NLP to unstructured financial texts, it becomes possible
to convert qualitative information into quantifiable
features that can be integrated into risk models. For
instance, NLP can identify subtle shifts in corporate
communication tone, detect mentions of adverse
events, or gauge market reactions to specific
announcements, providing early warning signals for
potential risks.
Machine Learning (ML) complements NLP by providing
the analytical horsepower to identify complex patterns,
make predictions, and classify risks based on the
features extracted from both structured and
unstructured data. From traditional ML algorithms like
Support Vector Machines (SVMs) and Random Forests to
advanced deep learning architectures such as Recurrent
Neural Networks (RNNs) and Transformers, ML models
can learn intricate relationships between various risk
factors and financial outcomes. When combined with
NLP, ML can build predictive models for credit default,
market volatility, operational failures, or even systemic
risk, offering a more dynamic and proactive approach to
risk management.
This paper aims to present a comprehensive framework
for
automating
financial
risk
assessment
by
synergistically com- bining NLP and ML. We will delve
into the various sources of unstructured financial data,
detail the NLP techniques employed for feature
extraction, and discuss the machine learning models
best suited for integrating these features
into a robust risk assessment system. Our objective is to
demonstrate how this integrated approach can provide
a more holistic, timely, and accurate understanding of
financial risks, thereby enhancing decision-making for
financial institutions, investors, and regulators. The
ultimate goal is to move to- wards an intelligent,
adaptive risk assessment system that can continuously
monitor, analyze, and predict financial risks in an
increasingly complex and interconnected global
economy.
2
Related Work
The intersection of financial risk assessment, Natural
Lan- guage Processing (NLP) and Machine Learning (ML)
has been a fertile ground for research, driven by the
increasing availability of unstructured financial data and
advancements in AI methodologies. Early work in
financial risk assessment primarily focused on
quantitative models, leveraging struc- tured financial
statements and market data to predict defaults, assess
creditworthiness, or forecast market volatility. These
models, often rooted in statistical methods like
regression analysis or econometric models, provided
foundational in- sights but were inherently limited by
their inability to process qualitative information.
The emergence of NLP marked a significant shift,
enabling researchers to tap into the rich qualitative data
The American Journal of Interdisciplinary Innovations and Research
68
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
embedded in financial texts. Initial applications of NLP in
finance in- volved sentiment analysis of financial news
and social media to predict stock market movements.
Loughran and McDon- alds` seminal work [
] highlighted
the importance of develop- ing finance-specific
sentiment lexicons, demonstrating that general-purpose
lexicons were inadequate for the nuanced language of
financial disclosures. Subsequent research ex- panded to
analyzing earnings call transcripts, analyst reports, and
regulatory filings, using techniques such as topic model-
ing to identify emerging themes and named entity
recognition to extract key financial entities.
With the rise of deep learning, NLP capabilities in finance
have advanced significantly. Pre-trained language
models like BERT [
] and its domain-specific
adaptations, such as Fin- BERT [
], have revolutionized
text understanding, enabling more accurate sentiment
analysis,
question
answering,
and
information
extraction from complex financial documents. These
models can capture intricate contextual relationships
and semantic meanings, leading to more robust feature
rep- resentations for downstream ML tasks. For
instance, deep learning models have been applied to
identify fraudulent ac- tivities from financial reports,
predict corporate bankruptcies, and assess the risk of
loan defaults based on textual loan applications.
Machine Learning, in parallel, has evolved from
traditional algorithms to sophisticated deep learning
architectures, of- fering powerful tools for pattern
recognition and prediction. In financial risk assessment,
ML models have been used to build credit scoring
systems, predict market crashes, iden- tify anomalies in
trading behavior, and optimize portfolio risk. The
integration of ML with NLP has created synergistic
opportunities. For example, features extracted from
finan- cial news sentiment (NLP) can be fed into a
predictive ML model to forecast market volatility (risk
assessment). Simi- larly, NLP-derived insights from
corporate disclosures can
enhance the accuracy of credit risk models that
traditionally rely solely on numerical data.
While significant progress has been made, existing liter-
ature often focuses on specific risk types or individual
data sources. For instance, many studies address credit
risk us- ing structured data and some textual features, or
market risk using news sentiment. However, a
comprehensive frame- work that seamlessly integrates
diverse unstructured textual data sources, processes
them with advanced NLP, and feeds them into a unified
ML pipeline for holistic financial risk assessment across
multiple risk categories remains an area requiring
further exploration. Our work aims to bridge this gap by
proposing such a comprehensive, automated frame-
work, leveraging the latest advancements in NLP and ML
to provide a more integrated and dynamic view of
financial risks.
3
Data Sources and Preprocessing
Effective financial risk assessment using NLP and ML
hinges on the quality and diversity of the data sources
utilized. Our framework integrates information from
both structured and, more critically, unstructured
textual data. The preprocessing steps are designed to
transform this raw, heterogeneous data into a format
suitable for machine learning models.
3.1 Unstructured Textual Data Sources
We consider a comprehensive set of publicly available
un- structured textual data sources that are highly
relevant to financial risk assessment:
•
Financial News Articles:
Sourced from reputable
finan- cial news outlets (e.g., Reuters, Bloomberg, Wall
Street Journal). These provide real-time insights into
market events, company-specific news, macroeconomic
indica- tors, and geopolitical developments. News
sentiment can be a leading indicator of market shifts.
•
Company Reports and Filings:
This includes
annual reports (10-K), quarterly reports (10-Q), earnings
call transcripts, and proxy statements. These
documents, available from the SEC EDGAR database,
offer detailed insights into a companys` financial health,
operational
strategies,
and
risk
factors.
The
Managements` Dis- cussion and Analysis (MD&A)
section, in particular, contains qualitative assessments
of a companys` perfor- mance and future outlook.
•
Social Media Data:
Platforms like Twitter (now X)
and financial forums (e.g., StockTwits, Reddits`
WallStreet- Bets) can provide a pulse on retail investor
sentiment, emerging trends, and rapid dissemination of
information (or misinformation). While noisy, this data
can capture collective market psychology.
•
Analyst Reports:
Reports from financial analysts
offer expert opinions, forecasts, and detailed company
valua- tions. These reports often synthesize complex
informa- tion and can influence institutional investor
behavior.
The American Journal of Interdisciplinary Innovations and Research
69
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
Regulatory Announcements and Central Bank State-
ments:
Official communications from regulatory bodies
(e.g., Federal Reserve, FDIC, ECB) and central banks
provide crucial information on monetary policy, regula-
tory changes, and economic outlook, directly impacting
systemic risk.
3.2 Text Preprocessing Techniques
Raw textual data is inherently noisy and requires
extensive preprocessing to extract meaningful features.
Our preprocess- ing pipeline involves several NLP
techniques:
•
Tokenization:
Breaking down text into individual
words or subword units (tokens). This is the first step in
converting raw text into a sequence that can be pro-
cessed by NLP models.
•
Stop Word Removal:
Eliminating common words
(e.g., "the", "is", "a") that carry little semantic meaning,
re- ducing dimensionality and noise.
•
Lemmatization/Stemming:
Reducing words to
their base or root form (e.g., "running" to "run",
"investing" to "invest"). This helps in standardizing
vocabulary.
•
Part-of-Speech (POS) Tagging:
Identifying the
gram- matical role of each word (e.g., noun, verb,
adjective). This can be useful for feature engineering.
•
Named Entity Recognition (NER):
Identifying and
classifying named entities in text into predefined cate-
gories such as person names, organizations, locations,
monetary values, and dates. In finance, this is crucial for
extracting company names, financial instruments, and
key figures.
•
Dependency Parsing:
Analyzing the grammatical
struc- ture of sentences to identify relationships
between words. This can help in understanding complex
financial statements.
•
Text Normalization:
Handling variations in text,
such as converting numbers to words, standardizing
dates, and correcting common misspellings.
3.3 Feature Engineering from Textual Data
Once preprocessed, textual data can be transformed
into nu- merical features suitable for ML models. Key
feature engi- neering techniques include:
Bag-of-Words (BoW) and TF-IDF:
Representing text as a
collection of words, with TF-IDF (Term Frequency-
Inverse Document Frequency) weighting words based
on their importance in a document relative to a corpus.
While simple, these are effective for many tasks.
Word Embeddings (Word2Vec [
], Fast-
Text):
Dense vector representations of words that cap-
ture semantic relationships. Words with similar mean-
ings are mapped to similar vectors. These are crucial for
capturing nuances in financial language.
Contextualized Embeddings (BERT [
], RoBERTa,
FinBERT [
]):
Advanced embeddings that generate word
representations based on their context within a
sentence. FinBERT, specifically fine-tuned on financial
corpora, is particularly effective for financial sentiment
and domain-specific understanding.
Sentiment Scores:
Deriving sentiment scores (positive,
neutral, negative) from text using lexicon-based meth-
ods or supervised learning models. These scores can be
aggregated at the sentence, paragraph, or document
level.
Topic Models (LDA [
], NMF):
Identifying latent top- ics
within a corpus of documents. This can help in
categorizing financial documents and understanding
the- matic risks.
Readability Scores:
Metrics like Flesch-Kincaid or
Gunning Fog Index can assess the complexity of finan-
cial disclosures, which may correlate with transparency
or obfuscation.
Lexical Features:
Counting specific types of words (e.g.,
negative words, uncertainty words, strong/weak modal
verbs) from finance-specific dictionaries.
3.4 Integration with Structured Data
For a holistic risk assessment, the features extracted
from unstructured text are integrated with traditional
structured financial data. This includes:
•
Financial Ratios:
Liquidity ratios, solvency
ratios, prof- itability ratios, and efficiency ratios derived
from finan- cial statements.
•
Market Data:
Stock prices, trading volumes,
volatil- ity, and macroeconomic indicators (e.g., GDP
growth, inflation rates, interest rates).
•
Credit Ratings:
Ratings from agencies like S&P,
Moodys`, and Fitch.
This integrated dataset, comprising both numerical and
NLP-derived features, forms the input for the machine
The American Journal of Interdisciplinary Innovations and Research
70
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
learn- ing models, enabling a more comprehensive and
nuanced risk assessment.
4
Methodology
Our framework for automating financial risk assessment
in- tegrates NLP for feature extraction from
unstructured data and ML for predictive modeling. The
methodology is de- signed to be modular, allowing for
flexibility in incorporating various NLP and ML
techniques based on the specific risk assessment task.
4.1 NLP Models for Feature Extraction
We leverage state-of-the-art NLP models to transform
raw textual data into actionable features:
Figure 1: Proposed Architecture for Automated Financial Risk Assessment
Sentiment Analysis:
For financial news, social media,
and company reports, we employ a fine-tuned FinBERT
model to generate sentiment scores (positive, neutral,
negative). This model is particularly adept at
understand- ing the subtle sentiment nuances in
financial contexts. The output is a continuous sentiment
score or a proba- bility distribution over sentiment
classes.
Topic Modeling:
Latent Dirichlet Allocation (LDA) [
] or
Non-negative Matrix Factorization (NMF) is applied to
large corpora of financial documents (e.g., 10-K fil- ings)
to identify underlying thematic risks. The output is a
topic distribution for each document, indicating its
relevance to various risk categories (e.g., operational
risk, market risk, credit risk).
Named Entity Recognition (NER) and Relation Ex-
traction:
Custom NER models, trained on financial
datasets, are used to extract key entities such as
company names, financial products, key personnel, and
dates. Re- lation extraction further identifies
relationships between these entities (e.g., "Company X
announced acquisition of Company Y"). This provides
structured insights from unstructured text.
Text Summarization:
For lengthy documents like reg-
ulatory filings, abstractive or extractive summarization
techniques are used to generate concise summaries,
high- lighting critical information relevant to risk
assessment.
4.2 Machine Learning Models for Risk Predic- tion
The features extracted from NLP, combined with
structured financial data, serve as input to various ML
models, chosen based on the nature of the risk and the
The American Journal of Interdisciplinary Innovations and Research
71
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
desired interpretability.
•
Credit Risk Assessment (e.g., Default Prediction):
Traditional ML:
Logistic Regression, Support Vector
Machines (SVMs) [
] are
effective for binary classification (default/non-default).
Features include financial ratios, credit ratings, and NLP-
derived sentiment scores from loan applications or
company news.
Deep Learning:
For more complex patterns, espe- cially
with sequential data (e.g., time-series of fi- nancial
news), Recurrent Neural Networks (RNNs) or
Transformer-based models can be used to pre- dict
default probabilities over time.
•
Market Risk Assessment (e.g., Volatility
Prediction):
Time-Series Models:
ARIMA, GARCH mod- els, or more
advanced deep learning models like LSTMs [
] or
Transformers are used to forecast market volatility. NLP-
derived sentiment from news and social media can serve
as crucial exoge- nous variables.
Ensemble Methods:
Gradient Boosting Machines
(GBMs) like XGBoost [
] or LightGBM can combine
various features (market data, macroeco- nomic
indicators, NLP sentiment) to predict mar- ket
movements or volatility spikes.
•
Operational Risk Assessment (e.g., Fraud
Detection):
Anomaly Detection:
One-Class SVMs, Isolation Forests,
or Autoencoders can identify unusual pat- terns in
operational data or textual reports that might indicate
fraud or operational failures. NLP features from internal
incident reports or employee communications can be
critical here.
Classification Models:
For known fraud types,
supervised classification models (e.g., SVM, Ran- dom
Forest, Neural Networks) can be trained on labeled
datasets, incorporating NLP features from suspicious
transactions or documents.
•
Liquidity Risk Assessment:
Regression Models:
Predicting future cash flows or
liquidity shortfalls using features from financial
statements, market conditions, and NLP-derived
sentiment about economic outlook or company- specific
news.
4.3 Model Integration and Deployment
The framework emphasizes a continuous learning and
adap- tive approach:
•
Feature Store:
A centralized repository for all pro-
cessed and engineered features (both NLP-derived and
structured) to ensure consistency and reusability across
different risk models.
•
Model Monitoring:
Continuous monitoring of
model performance (e.g., accuracy, precision, recall, F1-
score) and data drift. Retraining mechanisms are in place
to adapt to changing market conditions or data
characteris- tics.
•
Interpretability and Explainability (XAI):
For
critical financial decisions, understanding
why
a model
makes a certain prediction is crucial. Techniques like
SHAP (SHapley Additive exPlanations) [
] values or
LIME (Local Interpretable Model-agnostic Explanations)
[
] are employed to provide insights into feature
importance
and model behavior, especially for black-box deep learn-
ing models. This helps risk managers trust and validate
the automated assessments.
•
Automated Reporting and Alerting:
The system
gen- erates automated risk reports and alerts based on
prede- fined thresholds, enabling timely intervention by
human risk managers. This reduces manual effort and
speeds up response times.
5
Results and Discussion
While this paper proposes a conceptual framework, the
effi- cacy of integrating NLP and ML for financial risk
assessment has been demonstrated across numerous
studies. Our ap- proach, by combining diverse data
sources and advanced techniques, is expected to yield
superior performance com- pared to traditional
methods. We anticipate improvements in several key
areas:
•
Enhanced Accuracy:
By incorporating rich, qualitative
information from unstructured text, the models can cap-
ture nuances and early warning signals often missed by
purely quantitative approaches, leading to more
accurate predictions of risk events.
•
Timeliness:
Automated NLP and ML pipelines can
process vast amounts of data in near real-time,
providing timely risk assessments that enable proactive
decision- making, especially in fast-moving markets.
The American Journal of Interdisciplinary Innovations and Research
72
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
•
Holistic View:
The integration of diverse data sources
(news, reports, social media) provides a more compre-
hensive and holistic view of an entitys` risk profile, mov-
ing beyond siloed assessments.
•
Early Warning Signals:
NLPs` ability to detect subtle
shifts in sentiment, tone, or emerging topics can serve
as early indicators of potential risks, allowing for earlier
intervention.
•
Scalability:
Automated systems can process and an-
alyze data at a scale impossible for human analysts,
making them suitable for large financial institutions.
5.1 Quantitative Results (Illustrative Exam- ple)
To illustrate the potential impact, consider a
hypothetical credit risk assessment scenario. A
traditional model relying solely on financial ratios might
achieve an AUC (Area Under the Receiver Operating
Characteristic Curve) of 0.75. By integrating NLP-derived
features such as sentiment scores from company news
and management discussion analysis, the AUC could
potentially increase to 0.85 or higher. Similarly, for
market volatility prediction, incorporating sentiment
from social media and news could significantly improve
the R- squared value of time-series forecasting models.
5.2 Qualitative Insights
Beyond numerical improvements, the framework offers
sig- nificant qualitative benefits. For instance, NLP can
identify specific reasons behind a companys`
deteriorating creditwor- thiness from news articles, such
as supply chain disruptions or management scandals,
providing actionable insights rather than just a default
probability. In operational risk, NLP can help categorize
and prioritize internal incident reports, identi- fying
systemic weaknesses that might otherwise go
unnoticed. This level of detail and context is invaluable
for human risk managers.
6
Conclusion and Future Work
This paper has presented a comprehensive framework
for
automating
financial
risk
assessment
by
synergistically inte- grating Natural Language Processing
and Machine Learning. We have highlighted the critical
role of unstructured textual data in providing a more
holistic, timely, and accurate un- derstanding of financial
risks. By leveraging advanced NLP techniques for feature
extraction and sophisticated ML mod- els for prediction,
our proposed framework offers a robust solution to
enhance traditional risk assessment methodolo- gies.
Our approach addresses the limitations of purely
quantita- tive models by incorporating qualitative
insights from diverse textual sources, enabling the
detection of subtle signals and emerging risks. The
frameworks` modularity allows for con- tinuous
adaptation and improvement, integrating new data
sources and advanced algorithms as they emerge. The
empha- sis on interpretability ensures that the
automated assessments can be understood and trusted
by human experts, facilitating better decision-making.
6.1 Future Work
Several promising avenues for future research and
develop- ment exist:
•
Multimodal
Integration:
Extending
the
framework to incorporate other modalities, such as
audio (e.g., tone of voice in earnings calls) or visual data
(e.g., charts in reports, facial expressions in video
conferences),
could
provide
an
even
richer
understanding of financial senti- ment and risk.
•
Real-time Risk Monitoring:
Developing and
optimiz- ing the framework for real-time data ingestion
and con- tinuous risk monitoring, enabling immediate
alerts for critical risk events.
•
Causal Inference:
Exploring causal inference tech-
niques to move beyond correlation and understand the
true causal relationships between textual features,
finan- cial events, and risk outcomes.
•
Ethical AI and Bias Mitigation:
Investigating and
mit- igating potential biases in NLP and ML models,
particu-
larly
concerning
fairness
and
representativeness across different demographic or
company sizes, to ensure equi- table risk assessments.
•
Reinforcement Learning for Adaptive Strategies:
Ap- plying reinforcement learning to develop adaptive
risk management strategies that can learn and optimize
re- sponses to evolving market conditions and risk
profiles.
•
Explainable AI for Complex Models:
Further
research into advanced XAI techniques specifically
tailored for deep learning models in financial risk
assessment to provide more granular and intuitive
explanations for complex predictions.
By pursuing these directions, the integration of NLP and
ML in financial risk assessment can continue to evolve,
offer- ing increasingly sophisticated, intelligent, and
reliable tools for navigating the complexities of the
The American Journal of Interdisciplinary Innovations and Research
73
https://www.theamericanjournals.com/index.php/tajiir
The American Journal of Interdisciplinary Innovations and Research
global financial land- scape.
References
1.
T. Loughran and B. McDonald, “When is a liab
ility
not a liability? textual analysis, dictionaries, and 10-
ks,”
The Journal of finance
, vol. 66, no. 1, pp. 35
–
65,
2011.
2.
J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova,
“Bert: Pre
-training of deep bidirectional transform-
ers for language understandin
g,”
arXiv preprint
arXiv:1810.04805
, 2018.
3.
D. Araci, “Finbert: Financial sentiment analysis with
pre-
trained language models,”
arXiv preprint
arXiv:1908.10063
, 2019.
4.
T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Effi
-
cient estimation of word representations in vector
space,”
arXiv preprint arXiv:1301.3781
, 2013.
5.
J. Pennington, R. Socher, and C. D. Manning, “Glove:
Global vectors for word representation,” pp. 1532–
1543, 2014.
6.
D. M. Blei, A. Y. Ng, and M. I. Jordan, “Latent dirich
-
let allocation,”
Journal of machine Learning
research
, vol. 3, no. Jan, pp. 993
–
1022, 2003.
7.
S. Huang, N. Cai, P. P. Pacheco, S. Narrandes, Y.
Wang, and W. Xu, “Applications of support vector
machine (svm) learning in cancer genomics,”
Cancer
genomics & proteomics
, vol. 15, no. 1, pp. 41
–
51,
2018.
8.
L. Breiman, “Random forests,”
Machine learning
,
vol. 45, no. 1, pp. 5
–
32, 2001.
9.
S. Hochreiter and J. Schmidhuber, “Long short
-term
memory,”
Neural computation
, vol. 9, no. 8, pp.
1735
–
1780, 1997.
10.
T. Chen and C. Guestrin, “Xgboost: A scalab
le tree
boosting system,” pp. 785–
794, 2016.
11.
S. M. Lundberg and S.-
I. Lee, “A unified approach to
interpreting model predictions,” pp. 4765–
4774,
2017.
12.
M. T. Ribeiro, S. Singh, and C. Guestrin, “" why
should i trust you?" explaining the predictions of any
classifier,”
pp. 1135
–
1144, 2016.
