Mukesh Chinta
@vrsiddhartha.ac.in
Assistant Professor, Computer Science and Engineering
V R Siddhartha Engineering College
RESEARCH INTERESTS
Computer Networks
Scopus Publications
Scopus Publications
V. Kusumaniswar, Mukesh Chinta, and Sk. Nawaz Shareff
IEEE
Stacking Classifier is an ensemble learning technique in machine learning where the outcome of various base classifiers is forwarded as an input for meta classifier which makes the final decision. In any software industry a triage team will allocate the bugs raised in the software to the designated developers, nevertheless, there is a potential that they may give the bug to the incorrect category of developers, wasting time and effort. It will be simple for the triager to allocate the issue to the appropriate developer if he is able to accurately identify the kind of bug. Also studying the complete bug report and identifying the type of bug is time taking process for the developer. This system identifies the variety of bug using machine learning approach. The primary input to differentiate the type of bug is the description written for the bug report. The description is fed to the natural language processing (NLP) algorithm as an input for text processing, which performs operations like word tokenization, word lemmatization and converting text into numerical representation. BERT (Bidirectional Encoder Representations from Transformers) model is utilized to convert text into numerical data. This system uses Random Forest, Gradient Boosting and Support Vector Machine (SVM) as base classifiers and Logistic Regression as meta classifier. In this system, we can determine 3 kinds of bugs namely Concurrency, Semantic and Memory bugs using Stacking classifier. This paper presents a machine learning model which provides high precision than existing systems which utilizes other models like Decision Tree, Random Forest and Naive Bayes individually.
V. Kusumaniswari, Mukesh Chinta, and Sk. Nawaz Shareff
IEEE
Support Vector Machine (SVM) is a most widely used classification technique in machine learning as it gives a better accuracy than most of the existing techniques. These days, users frequently run across glitches and malfunctions when using software or applications. A lot more understanding is required by the developer to identify the type of bug, which takes time. In this system, to find the type of a bug a machine learning classification technique SVM. Here, bug report is fed as an input to a natural language processing algorithm. In this method, a dataset containing various types of bug reports is used to analyse the description in the given bug report using text processing and keyword extraction in natural language processing. This model can identify three types of bugs: memory bugs, concurrency bugs and semantic bugs using SVM. This system proposes a SVM model that gives better accuracy than the existing models which use several other techniques like decision tree, random forest and naive bayes.
Venkata Reddy Venna, Mukesh Chinta, Hari Sai Babu Avvaru, and Kushal Kumar Chintakayala
IEEE
Recent observations post covid have led us to believe that a way to reuse some of the goods owned by the students like books, electronic gadgets, and vehicles would be a good way to reduce the wastage and also the academic expenditure for the users. Our android application provides a platform for the students to post information about the goods that they wish to sell. Prospective buyers (only from our campus) can check the app, and then contact the seller if they are interested. No support or payment gateway is provided in this application for doing the financial transaction, but it provides a way to communicate and take further action. This platform not only focuses on the market but also helps the registered users to interact with the chatbot where they can clarify their doubts and issues regarding the product. When it comes to security, this application is strictly restricted to the members of the VRSEC community, where one can register with their college mail id and they also have to link their aadhar card and student identity card. An admin would be monitoring the whole process and can take necessary action if illegal/abusive content is displayed. Android Studio 4.2 with the help of java and kotlin is used for developing this application in major development languages.
Mukesh Chinta, Anagani Likhita, and Yamini Aravapalli
IEEE
Heavy rainfalls leading to floods in cities and villages is a common sight in our country. These situations lead to destruction of roadways and bridges, and often public infrastructure as an aftermath. Inspection of such facilities to assess the damage and identify any potential vulnerability is a tedious process. Some of the cracks/crevices might not be even visible to the naked eye. An automated system which can detect cracks saves money, time and even lives. This will help us improve road safety which is the reason for major accidents. The proposed work uses machine learning concepts to implement such a system which automatically detects the cracks on the roads, bridges and will send an alert to the concerned authorities there by potentially reducing the risk for disaster occurrence. Convolutional Neural Networks (CNN) can be used for the identification of cracks. By integrating the CNN Classifier with the camera, the cracks can be automatically detected in that region and reported.
Jasthi Siva Sai, Mukkamala Namitha, Routhu Ramya Dedeepya, Mulugu Suma Anusha, Angadi Lakshmi, and Mukesh Chinta
Springer Nature Singapore
Suma Anusha Mulugu, Namitha Mukkamala, Bhavana Gampa, and Mukesh Chinta
IEEE
In order to improve healthcare management, it is important to track health outcomes. For medical researchers, machine learning techniques have become a popular means of making accurate predictions. Machine learning techniques can identify trends in large amounts of dataset, and capable of predicting cancer accurately. In the current medical research, medical imaging plays a significant role in thorough examination and diagnosis of the entire human body. Medical professionals rely entirely on computed tomography results acquired from the image sensors. Lungs are the most primary organ of the human respiratory system. Medical professionals face a challenging task in accurately predicting the lung cancer. Detecting the lung cancer aids in determining the appropriate treatment, which increases the chances of survival for the lung cancer patient. In this project the lung cancer dataset is taken as input, the cancerous images are divided into five clusters based on the features obtained using K-Means clustering. Now, detection of cancer cells can be done by using transfer learning. In this project Densenet121 model in transfer learning is used to identify whether the cancer cells are present or not. If the CT scan image is cancerous then the treatment associated with the type is suggested to the user. This would help the physician to guide the patient on whether to take surgery or to take other kind of treatments. This is also used for Insurance companies. In this project, Image processing, transfer learning, and K-means clustering are used to cluster the cancer images based on features and to identify whether the image is cancerous or not and to suggest treatment.
Vijaya N, Mukesh Chinta, Kavya E, and Sumanth Varma L
IEEE
Pneumonia is a respiratory infection resulting in inflammation of the Lungs. Viruses, bacteria, or fungi could be to blame for this infection's sickness. Pneumonia can range in intensity from non-threatening to fatal. The most vulnerable demographics include infants and young children, seniors, those with health issues, and people with weakened immune systems. Early detection of pneumonia disease is essential for ensuring curative care and boosting survival rates. The most common approach for diagnosing pneumonia is a chest x-ray. However, the examination of chest radiographs is a subjective and difficult task. In this study, Convolutional Neural Network, Resnet, Inception Net was used for the analysis, as a result by comparing the above-mentioned models we got the accuracy more than 90% for Inception Neural Network. By these results new model has come up with the accuracy of the more than 90%. Future work includes the comparison of other models, among them Models with highest accuracy is chosen to design and develop the Dash Stream Application for the prediction of Pneumonia using X-ray images.
Mukkamala Namitha, Mulugu Suma Anusha, Gampa Bhavana, and Mukesh Chinta
IEEE
Operative mortality rates are a problem of great interest among surgeons, patients, because postoperative complications are the foremost reason for any form of thoracic surgery. The statistical optimization and probabilistic approaches used in the branch of artificial intelligence enables computers to “learn” from previous data and detect complicated patterns in large, noisy, or complex data sets. There are many machine learning methods used to predict the mortality of a patient after the lung cancer surgery. The data is collected from patients who underwent major surgeries like Heart Transplant, Lung Transplant and removal of parts of the lungs full of the cancer, this data is used as reference to predict the risk to the patient after the surgery. In this project the overall analysis is done by taking the patient's past medical records, daily habits and predicts outcomes based on records from previous year's surgeries. So, in our project we are building two models using Random Forest, SVM and then the model with best accuracy is used to predict the severity of patient after lung cancer surgery. This outcome will help the doctors to guide the patient on whether to have surgery or not. If doctors believe the surgery may impair the patient's quality of life and there is a known high probability of death within a year, then both parties can decide whether to follow through on surgery or decide an alternative treatment method. So, we will classify the post-operative life span of a patient into two classes i.e., high risk factor with chance of death after surgery and the other one is survival. Here, Random Forest, SVM, and Logistic Regression are used to predict the risk factor.
Rupa Ch., Sumaiya Shaikh, and Mukesh Chinta
IGI Global
In current days, there is a constant evolution in modern technology. The most predominant usage of technology by society is the internet. There are many ways and means on the internet through which data is transmitted. Having such rapid and fast growth of communicating media also increases the exposure to security threats, causing unintellectual information ingress. Steganography is the main aspect of communicating in an aspect that hides the extent of communication. Steganalysis is another essential concern in data concealing, which is the art of identifying the existence of steganography. A framework has been designed to identify the concealed data in the multimedia file in the proposed system. This work's main strength is analyzing concealed data images without embedding and extracting the image's payloads. A quantitative steganalysis approach was considered to accomplish the proposed objective. By using this approach, the results were achieved with 98% accuracy.
Hema Pallevada, Siva parvathi Potu, Teja Venkata Kumar Munnangi, Bharath Chandhra Rayapudi, Sai Raghava Gadde, and Mukesh Chinta
IEEE
Agriculture is the backbone of India. Fertilizers play a key role in the agricultural yield. A key problem faced by the farmers is lack of knowledge on the amount of fertilizers to be used. Farmers think that higher the fertilizer used, greater the productivity. But it is not correct, the soil uses the exact amount it needs and leaves the rest. Over utilization leads to leaching and decrease in the natural soil fertility and many such problems. A solution is provided by allowing the farmers to test their lands and use the fertilizer as per the soil’s need at an affordable cost. This work gives a report about the design of cost efficient soil nutrients detection using pre-prepared capsules. Here test can be performed for three different types of nutrients Sodium, Potassium and Phosphorous. Here three test tubes are taken and each one is filled with certain amount of soil and water, and then the mixture is shaken for 15 minutes. Then there occurs a color change in the tube. Here a color sensor is used and the color change in the test tubes is detected by the sensor and compared with the existing information about color-deficiency. Sensory data is processed using Arduino and then information about the deficiency and amount of fertilizer needed to overcome the deficiency is given to the farmer.
Hema Pallevada, Gayathri Phani Kumar Kanuri, Sravani Posina, Satyendra Paruchuri, and Mukesh Chinta
IEEE
Online vehicle booking service became a growing need now-a-days as it offers an efficient and cheaper alternative. One can directly call a taxi to any location at any time they want without the need for a lengthy delay. There are various platforms implementing online vehicle booking service but the major issue with them is having a centralized authority which are pocketing huge profits. A fixed percentage of money paid by the user is taken by the centralized authority. These centralized systems also tend to misuse their user’s data and are also prone to several attacks. One of the best solutions to this problem is to use BlockChain. Our objective is to avoid this centralized authority by creating a decentralized application. Not just removing the central authority, BlockChain also provides several features like immutability, better transparency, enhanced security and traceability. Platforms like Ethereum and Hyperledger allow one to develop decentralized applications and deploy them in the blockchain. Smart Contracts can be developed using Solidity to achieve decentralization. This maintains transparency, immutability and removes a central authority which may misuse its user's information.
Valaparla Rohini, M. Sobhana, Ch. Smitha Chowdary, Mukesh Chinta, and Deepa Venna
Springer International Publishing
Ch.Raga Madhuri, Mukesh Chinta, and V V N V Phani Kumar
IEEE
Since the beginning, the fundamental goal of man is to make life easy to live. The whole world believes that wealth would make life comfortable and luxurious. One of the most common notion among humans is that one of the best way to make money is to invest in stock markets which are expected to have tremendous results. There is a requirement to develop an intelligent system to perform predictions based on various indicators like fundamental, statistical and technical trends. However, there is no one good predictive model that has been successful to beat the trends in market continuously. Traditionally for time series data, the predictions are in general performed based on past historical data and market trends, historical correlation data and projections can be calculated. Above all said, there is no such system that calculates the predictions based on users selection on investment type and on risk criteria user is willing to take. So in this paper, we tried to demonstrate the technique(s) to get most accurate results.