Research & Review: Machine Learning and Cloud Computing(e-ISSN:2583-4835)

Evaluation of Soil Quality for Crop Prediction Based on Feature Selection in Machine Learning

Dr. Rasina Begum, S Janarthanan, B Naveen, P Pandeeswaran — Sat, 15 Jul 2023 00:00:00 +0530

Farmers can practice an effective understanding of the soil's idiosyncrasies allowing for more crops to be grown with fewer resources. Soil prediction relies heavily on calcium, phosphorus, pH, and soil organic carbon. These traits have a substantial impact on crop productivity. The method employs two independent machine learning models, Using KNN and random forest regression; specific soil parameters can be predicted. Crop productivity is boosted as a result of accurate crop prediction. This is where machine learning in the field of crop prediction comes into play. Crop forecast is influenced by geographic, meteorological, and soil characteristics. An integral aspect of the prediction process used by feature selection techniques is choosing the proper features for the right crop or crops. This paper uses categorization approaches to recommend the appropriate crop or crops for the area and conducts a comparative study of multiple wrapper feature selection methods. According to the experimental findings, the adaptive bagging classifier and recursive feature elimination approach surpass the competition. Based on the Africa soil property prediction dataset, the performance of these models is assessed. Knowing the characteristics of the soil in their particular terrain will be useful to the farmers. This study investigates how effectively various machine learning approaches can predict soil qualities crucial to agriculture using spectroscopic data.

Exploring Machine Learning Techniques for Data Analysis in the RideSharing Industry

Rajendra Prasad R, M.S. Shashidhara — Wed, 26 Jul 2023 00:00:00 +0530

Uber, a popular ride-sharing platform, generates vast amounts of data that can be leveraged to gain valuable insights and optimize its operations. This research paper focuses on the application of machine learning techniques for data analysis in the context of Uber. The paper aims to provide a comprehensive overview of the existing literature and research on Uber-related data analysis using machine learning, covering various aspects such as demand forecasting, surge pricing optimization, rider behaviour analysis, driver allocation, and routing.

In the domain of demand forecasting, researchers have explored time series analysis, regression, and deep learning models to accurately predict ride demand and optimize driver allocation. Surge pricing optimization has been addressed through reinforcement learning algorithms, dynamic pricing models, and approaches that consider user behaviour and market dynamics. Rider behaviour analysis has employed machine learning techniques for customer segmentation, churns prediction, and personalized recommendations. Driver allocation and routing have been optimized using machine learning algorithms to efficiently allocate available drivers based on factors like trip duration, driver availability, and traffic conditions.

Additionally, privacy and security concerns associated with Uber data analysis have been addressed through privacy-preserving techniques and secure computation methods. The paper discusses the challenges faced in these areas and identifies potential future research directions.

Detecting Persecution on Interactive Networks Using Machine Learning Methods

Sheela B V, Rajesh N — Thu, 27 Jul 2023 00:00:00 +0530

Persecution on interactive networks, such as
social media platforms and online forums,
poses significant challenges to ensuring a safe
and inclusive digital environment. Detecting
and addressing instances of persecution is
crucial for safeguarding individuals' rights
and promoting online harmony. This research
paper presents a novel approach to detecting
persecution on interactive networks using
machine learning methods.
The proposed framework leverages the power
of machine learning algorithms to
automatically identify patterns and indicators
of persecution within user-generated content.
The process begins with the collection and
preprocessing of a large-scale dataset
comprising diverse interactions on interactive
networks. Various features, including textual,
contextual, and user-based attributes, are
extracted to capture the nuanced aspects of
persecution.
Next, a range of machine learning techniques,
such as natural language processing,
sentiment analysis, and social network
analysis, are employed to analyze the dataset.
Multiple classification models, such as
support vector machines, random forests, and
deep learning architectures, are trained and
evaluated to identify the most effective
approach for persecution detection.
The experimental results demonstrate the
effectiveness of the proposed methodology in
accurately detecting instances of persecution
on interactive networks. The trained models
achieve high precision and recall rates,
highlighting their ability to discern subtle
instances of persecution and distinguish them
from non-persecutory content. The findings of
this study contribute to the development of
automated tools that can aid in identifying
and mitigating persecution on interactive
networks, thereby fostering a safer and more
inclusive digital space.

A Review on Intelligent Malware Detection Using a Machine Learning Approach

Tiab Ali, Pradeep Kumar — Tue, 29 Aug 2023 00:00:00 +0530

In the present period, cell phones are
becoming famous with different applications
(applications) to make our lives more
straightforward. A few versatile Working
Frameworks (operating systems) are
accessible in the market including iOS,
Android, BlackBerry and Windows
Telephone. Android is a broadly utilized
portable operating system with a piece of the
pie of over 85%. It depends on the Linux
piece explicitly worked for touchscreen
gadgets like tablets cell phones and so on. In
the on-going time, there is an expansion in the
use of cell phones for different purposes like
banking, virtual entertainment, training and
so on. The developing prominence of Android
applications has baited aggressors to make
pernicious applications that represent a few
dangers, for example, monetary misfortune,
data spillage and so on. These pernicious
applications are turning out to be more
complex and utilizing better approaches to
target cell phones. These can avoid location
and relief strategies that have previously been
created. The customary security frameworks
like interruption identification/counteraction
frameworks and Against Infection (AV)
programming depend on signature-based
techniques and accordingly can't recognize
new-age malware. Hence, there is a need to
plan methods for better malware
distinguishing proof and grouping. Besides, in
a true situation, the quantity of tests shifts
considerably among different malware
families. In this manner, it is vital to assemble
malware arrangement models that can deal
with imbalanced classes. Moreover, there is
an absence of sufficient exploration to break
down the dangers or hazards presented by
Android applications. The fundamental point
of this examination is to resolve these issues
and give powerful arrangements. AI (ML)
procedures have been utilized to distinguish
malware given characteristics mined utilizing
static and dynamic malware examination.
Through tests, it is seen that the two sorts of
malware examinations have their upsides and
downsides. The obscure malware utilizes
progressed muddling procedures to conceal
its presence, and it can distinguish the
sandbox climate where it is running.
Subsequently, the single methodology either
static or dynamic can't recognize and
characterize obscure malware. A coordinated
methodology (a blend of static and dynamic
ascribes) has been proposed in this work
which can break down, recognize and group
the malware

Advancements in Activity Recognition Technologies

Prathiksha S Raj, R Ashok Kumar — Mon, 11 Sep 2023 00:00:00 +0530

In today's data-rich era, marked by the
widespread use of data collection tools like
smartphones and video cameras, Human
Activity Recognition (HAR) has become a
dynamic and versatile field with numerous
practical applications. The fusion of these
devices with artificial intelligence (AI) has
revolutionized our ability to detect and
understand human activities with
unparalleled precision, revealing hidden
insights.
The proliferation of electronic gadgets
equipped with sensors, cameras,
accelerometers, and gyroscopes has
significantly expanded the scope of HAR.
These devices serve as the bedrock for
collecting intricate data on human movements
and behaviours. Simultaneously, AI
advancements empower us to process and
interpret this data accurately, providing
profound insights into various human
activities.
HAR hinges on three interconnected pillars:
data acquisition devices, AI algorithms, and
diverse applications that leverage their
synergy. This paper conducts a
comprehensive exploration of these pillars,
drawing insights from extensive literature
and real-world datasets encompassing various
contexts and data sources.
Notably, neural networks play a pivotal role
in advancing HAR techniques. These AIdriven algorithms are the cornerstone of
precise activity recognition. The paper delves
into the intricacies of neural networks,
shedding light on their ability to reveal
concealed information and enable nuanced
activity interpretation.
Throughout our exploration, we confront the
challenges confronting HAR, including data
pre-processing, noise reduction, model
optimization, and generalization. Addressing
these challenges is essential for enhancing the
accuracy and applicability of HAR methods.